The evolutionary ecology of symbiotic ant-plant relationships

J. HYM. RES.
2(1), 1993 pp. 13-83

The Evolutionary Ecology of Symbiotic
Ant - Plant Relationships

Diane W. Davidson and Doyle McKey

(DWD) Department of Biology, University of Ulah, Salt Lake City. Utah 84112;

(DM) Department of Biology, Universily of Miami, Coral Gables. Florida 33124

Abstract . — A tabular survey of ant-plant symbioses worldwide summarizes aspects of the evolutionary ecology of
these associations. Remarkable similarities between ant-plant symbioses in disjunct tropical regions result from
convergent and parallel evolution of similarly preadapted ants and plants. Competition among ants has driven
evolutionary specialization in plant-ants and is the principal factor accounting for parallelism and convergence. As
habitat specialization accompanied the evolutionary radiation of many myrmecophytes, frequent host shifts and de
novo colonizations by habitat-specific ants both inhibited species-specific coevolution and co-cladogenesis, and
magnified the diversity of mutualistic partners.

The comparatively high species diversity of neotropical plant-ants and myrmecophytes probably results from
two historical factors. Most importantly, influenced by Andean orogeny, greater habitat disturbance by fluvial
systems has created a mosaic of habitat types unparalleled in other tropical regions; both myrmecophytes and plant-
ants have diversified across habitat boundaries. Second, the arrival of a new wave of dominant ants (especially
Crematogaster) may have condensed the diversity of relatively timid plant-ants to a greater degree in Africa and
Asia than in the more isolated Neotropics. Regular trajectories in the evolutionary histories of plant-ants appear
to be driven principally by competition, in a manner analogous to the taxon cycles or pulses proposed for other
groups.

“In all the plants I have seen bearing sacs on the leaves, to whatever order they belong, it is remarkable that the
pubescence consists of long hairs having a tubercular base; and although I do not see what connection that peculiarity
can have with the ants’ choice of a habitation, it is probable they find some advantage in it.” .... “Ants’ nests in
swellings of the branches are found chiefly in soft-wooded trees of humble growth, which have verticillate or quasi-
verticillate branches and leaves, and especially where the branches put forth at the extremity a whorl or fascicle of
three or more ramuli; then, either at each leaf-node or at least at the apex of the penultimate (and sometimes of the
ultimate) branches, will probably be found an ant-house, in the shape of a hollow swelling of the branch...” (Spruce
1908).

INTRODUCTION

A synthetic overview of the evolutionary ecol-
ogy of mutualism has been disappointingly slow to
develop (Bronstein 1991). In large part, this short-
coming may reflect the composite nature of
mutualisms, which often arise as parasitisms (Th-
ompson 1 982), and frequently convey benefits con-
tingent on physical environments, population den-
sities, and third or multi-species interactions (re-
viewed in Howe 1984, Addicott 1985, Law and
Koptur 1 986, Schemske and Horvitz 1 988, Thomp-
son 1988, Cushman and Addicott 1991). The lack
of a conceptual organization for such complex and
variable associations inhibits a search for patterns

in historical and ecological factors shaping the
evolution of mutualism. Complicating this en-
deavor still further is that most studies of mutualism
focus on pollination and dispersal systems, which
account for 80 % of the articles on mutualism in
Bronstein’ s ( 1991) survey. Despite excellent treat-
ments available for taxonomically and/or geographi-
cally restricted suites of such interactions (e.g.,
Heithausetal. 1975, Feinsinger and Colwell 1978,
Janson 1983, Herrera 1984, Gautier-Hion et al.
1985, Moermond and Denslow 1985, Gottsberger
1990, Bronstein 1992), both the overwhelming
numbers and the taxonomic and ecological diver-
sity of these interactions magnify the difficulty of

14

Journal of Hymenoptera Research

identifying single or few organizing processes or
principles.

Symbiotic associations between ants and
myrmecophytic plants offer a useful counterpoint.
Sufficiently small in number to be summarized in a
single table (Appendix 1 ), they nevertheless occur
in numbers adequate to provide fertile substrate for
hypothesis testing. Their presence in tropical re-
gions throughout the world facilitates comparisons
among taxonomic and ecological equivalents
evolved in isolation on different continents (McKey
and Davidson, in press). Despite their considerable
diversity and widespread distribution, these rela-
tionships are relatively uniform in structure. Thus
all myrmecophytic plants provide permanent hous-
ing and food to ants which are known or (more
often) presumed to protect their hosts from her-
bivory or competition, or to provision them with
nutrients ( reviewed recently in Beattie 1 985, Huxley
1986, Jolivet 1986, Holldobler and Wilson 1990).

Here we provide an overview of the symbiotic
ant-plant relationships, focusing principally on trees,
shrubs and hemiepiphytes of the American and
African tropics. (The epiphytic ant-plants have
been reviewed recently elsewhere by Davidson and
Epstein 1989.) This geographic specialization re-
flects our comparatively poor understanding of ant-
plant relationships in the Oriental and Australian
tropics where, with the exception of ant-epiphytes
(Jebb 1985, Huxley and Jebb 1991 ), investigations
are fewer in number and less detailed (but see the
recent proliferation of work by Fiala and Maschwitz
1990 and 1991, Fiala et al. 1989 and 1991,
Maschwitz et al. 1989 and 1991). For
myrmecophytes overall, existing evidence is often
too meagre for a convincing assessment of the
fitness consequences of particular associations. We
therefore avoid using the terms “mutualism” and
“facilitation” in favor of less restrictive words like
“association”, “interaction”, or “relationship”. For
similar reasons, the terms “myrmecophyte”,
“myrmecophytic” and “ant-plant” are used here
only to describe plants regularly inhabited by ants,
without implying that plants either benefit from the
ants or possess traits evolved principally as ant
attractants. On occasion, we also refer to
“myrmecophilic” plants, those which are not sym-
biotic with ants but produce obvious ant attractants

such as extrafloral nectaries (EFN’s) and/or pearl
bodies.

Our principal themes here are the factors which
have predisposed particular ants and plants toward
symbiotic association, and the ecological forces
which have driven evolutionary specialization in
each of these taxa. We also summarize the pro-
cesses generating and maintaining diversity within
each of these groups, as well as the factors limiting
species specificity and co-cladogenesis. Finally,
we speculate about particular evolutionary trajec-
tories which appear to have occurred regularly
across independent lineages of plant-ants and ant-
plants. As a prelude to all the above, we briefly
review the way in which historical context appears
to have influenced the evolution of ant-plant sym-
bioses in the American and African tropics.

DIVERSITY, BIOGEOGRAPHY AND HISTORY

Both plant-ants and myrmecophytes achieve
their greatest richness in the American tropics
(McKey and Davidson, in press). Among ants, the
proportion of neotropical and African genera con-
taining specialized plant-ants is approximately the
same, whether calculated by biogeographic region
(respectively, 10 % and 12 % of genera) or for
mesic tropical environments (12.8 % and 14.5 %,
respectively). Although the mesic Neotropics hold
approximately 1 .3 times as many ant genera as does
mesic tropical Africa (Brown 1973), the latter land
mass has slightly more genera which contain at
least one plant-ant. Nevertheless, two of these gen-
era are monotypic and, based on present knowl-
edge, the species richness of plant-ants appears to
be about 3.5-fold greater in the Neotropics than in
Africa (current estimates of 85 species versus 24,
including one species in Madagascar). Differences
in diversity occur principally due to the prolifera-
tion of plant-ant species within endemic neotropical
genera. In the New World, significant radiations of
plant-ants occur in endemic Pseudomyrmex
(N = 32 species), Azteca (N probably > 20),
Myrme/achista (N > 6), and Allomems (N 8), as
well as in cosmopolitan Pheidole (N 6) and
Pachycondyla (N 4). In contrast, significant
radiations of African plant-ants are limited to
Tetraponem (N 5) and Technomynnex (N 6),

Volume 2, Number 1, 1993

15

both widely distributed in the Old World tropics,
and even these radiations are comparatively small.

Relative to the ant faunas of both the American
and African tropics, those of the Oriental and Aus-
tralian regions appear to be poor in plant-ant genera
(McKey and Davidson, in press); respectively, only
5.6 % and 7.3 % of regional ant genera, and 7.3 %
and 9. 1 % of mesic tropical genera, contain plant-
ants. Moderate to large radiations of plant-ants in
the Oriental region include only cosmopolitan
Crematogaster{ N 8 species) and Compouotus ( N
7), as well as endemic Cladouiynua (N 5), and
current estimates of plant-ants are only 24 species
overall. In the Australian region, encompassing
northern Australia, New Guinea and associated
islands, such radiations are limited to endemic
Anonychomynna (probably > 3 species), and the
species richness of plant-ants presently stands now
at only about 1 2 species. Although the numbers of
plant-ants may increase slightly in these regions
due to increased sampling effort (cf. Dorow and
Maschwitz 1990, Maschwitz et al. 1991 ) and taxo-
nomic revision (e.g., S. Shattuck, 1 99 1 , 1 992b), the
relative poverty of plant-ants at the generic level is
likely real.

Myrmecophytes probably constitute a similar
fraction of all plant genera in the American and
African tropics, but their species richness is dis-
tinctly greater in the Neotropics (McKey and
Davidson, in press). Again unmatched in Africa,
major radiations of ant-plants within (mainly) en-
demic, neotropical genera largely account for this
difference. Neotropical plant genera with signifi-
cant radiations of myrmecophytes include endemic
Cecropia (N 50-60 ant-plant species), Tachigali
(N 20), Triplaris (N = 17), Tococa (N = 40-45),
Clidemia (N = 1 5-20) and Maieta (N 15), as well
as non-endemic Acflc/n(N 12 species), Ocotea( N
6) and Hirtella (N = 6). In contrast, in Africa only
Acacia (N 15) and, to a lesser extent, Cuviera
(N = 8+). Cambium (N = 3-6) and Clerodendrum
(N 3) contain moderate to large numbers of ant-
plants, and of these genera only Cuviera is re-
stricted to the Ethiopian region. Estimates of
myrmecophyte species richness are about three-
fold greater in the American than the African trop-
ics, and maximum local (alpha) diversity may be
twice as high. Although it is not yet possible to

estimate the frequency of myrmecophytes in the
tropical floras of Oriental and Australian regions,
substantial radiations of myrmecophytes within
genera are comparatively limited (references in
McKey and Davidson [in press]). These probably
include only Macaranga(N 23), Korthalsia{ N =
7+) and Neonauclea (N = 4+) in the Oriental
tropics, and Chisocheton (N = 6), Kibara,
Steganthera , and Semecarpus (each with N = 4) in
the Australian tropics. Altogether, the Oriental and
Australian tropics likely hold slightly more than
100 myrmecophyte species.

At the generic level, the determinants of ant-
plant and plant-ant diversity in the American and
African tropics are probably similar to those regu-
lating species richness of the floras and ant faunas
overall (McKey and Davidson, in press). Radia-
tions of myrmecophytes and plant-ants in both
areas appear to have been strongly affected by both
the climatic and geologic histories of the continents
and to have been correlated with diversification in
habitat use. As may be common for neotropical
plants in general (Gentry 1986, 1989 and in press,
but see Simpson and Todzia [1990] for the high
Andean flora), generic radiations of ant-plants may
often be comprised of neoendemics with compara-
tively recent origins. Frequently geographically or
edaphically restricted, such species may be prod-
ucts of a “species pump”, postulated to have gener-
ated new species through habitat specialization
during range reexpansions within interglacial inter-
vals of the Pleistocene (Colinvaux, in press). Al-
though the diversity of tropical ant species has not
previously been related explicitly to any similar
mechanism, a possible link between speciation and
habitat specialization is evidenced by the observa-
tion that many plant-ants show greater specificity
to habitats than to host species (Benson 1985,
Davidson etal. 1989 and 1991;Longino 1989a and
1991a).

Given historical and contemporary differences
in geological activity, and in correlated rates of
habitat disturbance on the two continents, the Ameri-
can tropics should have provided greater opportu-
nity than did tropical Africa for habitat specializa-
tion and speciation (McKey and Davidson, in press).
Topographically, the mesic African tropics occupy
a comparatively flat and featureless plain, much

16

Journal of Hymenoptera Research

more homogeneous than mesic tropica] America.
In the Neotropics, orogenic activity in the Andes
has not only influenced the montane and submontane
areas directly, but has given rise to the fluvial
disturbances that helped to create a spectacular
mosaic of landscapes over the vast Amazonian
region. No less than 26 % of modem lowland
forests of Western Amazonia give evidence of
recent erosional and depositional activity, and ap-
proximately 12 % of these lands are currently in
some stage of succession (Salo et al. 1 986, Rasanen
et al. 1 987). In addition to their role in creating and
maintaining a landscape mosaic conducive to rapid
speciation, the Andes also appear to have protected
the mesic Neotropics from the severe and frequent
droughts which could have magnified species ex-
tinctions in Africa, as mesic forests were repeatedly
reduced and fragmented during Pleistocene times
(Raven and Axelrod 1974, Axelrod and Raven
1978).

Finally, neotropical species should also have
received greater protection than their African coun-
terparts from Pleistocene temperature variation.
Lowland Africa is approximately 500m higher in
elevation than is lowland Amazonia, and would
have provided fewer refugia for plants and animals
during glacial periods. Current evidence (e.g.,
Bengo and Maley [1991]) indicates that montane
forest, including elements now restricted to the
cool, moist conditions of the Afromontane zone,
extended to low elevations (600 m or perhaps
lower) in Central Africa during several periods
over the last 135,000 years. Judging from the
dramatic drop in ant diversity and abundance with
elevation on humid tropical mountains (Janzen
1973), the conditions suggested for these periods
would not have been conducive to the success of
much of the contemporary ant fauna of lowland
African forests. To the extent that climatic fluctua-
tions in Africa exceeded those in the American
tropics, these could have led to the dissolution of
mutualisms, even without species extinctions, as
the fitness consequences of association shifted (e.g.,
to parasitism) with fluctuations in the abiotic and
biotic environments.

SIMILARITIES BETWEEN ANT-PLANT
RELATIONSHIPS OF DIFFERENT
TROPICAL REGIONS

In the context of the aforementioned differences
in species richness, and in the climatic and geologic
histories of ant-plants on different tropical land
masses, certain similarities in the form and ecology
of ant-plant relationships of different continents
appear all the more striking. For example, across
tropical land masses, large colonies of active and
aggressive ants occupy fast-growing and light-
demanding pioneer trees (neotropical Cecropia and
Old World Macaranga). In contrast, timid ants
inhabit small, slow-growing understory shrubs or
treelets with hairy domatia (e.g., American Hirtella,
Duroici, and many melastomes, and African
Magnistipula, Delpydora, Cola, and Scapho-
petalum). Finally, myrmecophytic trees of second-
ary forests and forest light gaps (neotropical Triplaris
and African Barteria) grow in circular clearings
made by pseudomyrmecine ants, which attack veg-
etation in the neighborhood of their hosts. McKey
and Davidson (in press) have amassed evidence
against common ancestry as a general explanation
for these remarkable commonalities. While some
comparisons between Africa and Asia suggest com-
mon descent of ant-plants, plant-ants or both,
myrmecophytes and specialized plant-ants appear
to have evolved largely independently in America
and Africa. No ant-plants of Africa and the
Neotropics have apparently shared a myrmecophytic
common ancestor. In contrast, the plant-ant habit
may be ancient in the sub-family Pseudomyrmecinae
and in tribes Myrmelachistini and Tapinomini, and
might possibly have preceded the splitting of South
America and Africa. However, with these possible
exceptions, resemblances between symbiotic asso-
ciations in the American and African tropics are not
due to common descent of one or both partners from
an association that predated continental separation
or other vicariance events, or which migrated intact
from one continent to the other (McKey and
Davidson, in press).

The remarkable correspondences between ant-
plant associations in the American and African
tropics must therefore be due to some combination
of: ( 1) parallel evolution of ants and/or plants from

Volume 2 , Number 1 , 1 993

17

similar starting material, (2) evolutionary conver-
gence, and (3) the matching of symbiotic partners
according to a set of shared rules. The task then is
to identify the preadaptations which have been
pressed into service and evolutionarily modified in
symbiotic ants and plants, and to recognize the
selection pressures which have led repeatedly to the
correspondences noted above.

PREADAPTATIONS OF PLANTS AND ANTS

Parallel and convergent evolution are usually
regarded as evidence that selection pressures have
acted in similar ways on organisms of different
lineages. Selection, however, is only part of the
explanation for these phenomena. Different lin-
eages may follow similar evolutionary trajectories
because they share similar developmental con-
straints which channel the action of selection along
a limited number of paths.

Preadaptations for Myrmeeophytism

The evolutionary antecedents of specialized
myrmecophytic traits are poorly explored. How-
ever, comparative studies of myrmecophytes and
their less specialized relatives are begining to sug-
gest plausible and testable hypotheses about the
origins of these traits (Benson 1985, McKey 1989
and 1991, O' Dowd and Willson 1989, Fiala and
Maschwitz 1991, Schupp and Feener 1991). For
example, in various plant taxa, a few similar struc-
tures have repeatedly provided the raw materials
transformed by selection into myrmecophytic struc-
tures. An understanding of the origins of these
traits may help to identify constraints which have
pressed ant-plants of diverse lineages and biogeo-
graphic regions into a limited number of molds. It
may also indicate developmental patterns which
have facilitated the evolution of myrmeeophytism,
and suggest why myrmecophytes have evolved
repeatedly in some lineages, but rarely or never in
others.

Provision of Food for Plants. — Discussion of
the evolutionary background of myrmecophytes
has tended to emphasize the provision of food for
ants. Indeed, there is evidence from many lineages
that the ancestors of ant-plants possessed extrafloral

nectaries, pearl bodies, or other traits, which pro-
vided food for ants in loose non-symbiotic interac-
tions. The large, complex nectary glands of some
ant-plants (e.g.. Acacia, Endospermum, and some
Macaranga ), and the elaborate food bodies of oth-
ers (e.g., Mullerian bodies of Cecropia, and
Beccarian bodies of Asian Macaranga) are readily
accounted for as outgrowths of these traits. As ant-
plant interactions intensified into symbiosis, such
attributes should have been easily modified by
selection acting on the composition and rate of
supply of food for ants. The Beltian bodies of
Central American ant-acacias may be the only case
in which a specialized food-producing structure of
a myrmecophyte lacks an obvious antecedent among
unspecialized but related plants.

Provision of food ensures that ants are a regular
component of the plant’s biotic environment, and
doubtless facilitates the evolution of more intense
interactions. However, myrmecophytes have
evolved in only a small subset of the numerous
plant lineages whose members are engaged in op-
portunistic myrmecophilic interactions; other plant
traits must also play a role in facilitating or con-
straining the evolution of symbiotic interactions.
Furthermore, in many cases, neither the
myrmecophytes nor their close relatives provide
food directly to ants. In many cases, EFN’s and
food bodies are lacking, and scale insects
(Coccoidea, Homoptera) are a major source of
colony nutrition (Appendix 1). Following Ward
( 1 99 1 ), we suggest that many myrmecophytic rela-
tionships evolved not from pre-existing
myrmecophilic relations, but from parasitisms in
which stem-nesting ants began to inhabit live plant
cavities and to tend Coccoidea.

Structures for Housing Ants. — We must thus
explore plant traits that facilitated the production of
cavities that could be modified by selection into
specialized structures for housing ants. The evolu-
tionary antecedents of myrmecodomatia, the defin-
ing feature of specialized myrmecophytes, have
received little attention. Preadaptations and devel-
opmental constraints in the evolution of
myrmecodomatia will be discussed in detail else-
where (McKey, in preparation) and are summa-
rized only briefly here.

18

Journal of Hymenoptera Research

Table 1 Taxa in which at least some myrmecophytes have long, dense hairs which inhibit insect movements on
stems, domatia or both.

Region

Family

Genus

ETHIOPIAN

Chrysobalanaceae

Magnistipiila

Dichapetalaceae

Dichapetalum

Ebenaceae

Diospyros

Rubiaceae

Canthium

Cuviera

Sapotaceae

Delpydora

Sterculiaceae

Cola

Scaphopetalinn

NEOTROPICAL

Boraginaceae

Cordia

Chrysobalanaceae

Hirtella

Fabaceae

Platymiscium

Tachigali'

Gesneriaceae

Besleria

Melastomataceae

Allomaieta 1 2
Blakea 3
Clidemia 4
Conostegia
Henriettea 5
Maieta
Sagraea 6
Toeoca 7

Cecropiaceae

Pourouma

Polygonaceae

Trip laris

Rubiaceae

Duroia 8

Hoffmannia

Remijia

ORIENTAL

Melastomataceae

Medinilla

Verbenaceae

Callicarjja

Piperaceae

Piper

AUSTRALIAN

Monimiaceae

Steganthera

1 At least one species, collected from a hillside over the junction of the Rio Sotileja and the Rio Manu, in southeastern Peru
(D. Davidson, unpublished).

Closely related to Maieta (A. Gentry, personal communication)

3 Benson ( 1 985) considers the leaf pouches of B. fonnicaria to be in transition from acarodomatia to ant-domatia. Among
the melastomes listed here, Blakea is unique in not belonging to the Miconieae.

4 At least three independent origins of domatia in Clidemia sensu strictu; includes Myrmidone (Judd and Skean 1991 )

s Includes Henriettella (Judd 1989)

6 Includes Ossaea p.p. (Judd 1989)

7 Includes Microphysca (Judd and Skean 1991)

8 Two independent origins of domatia (foliar domatia and swollen intemodes)

Volume 2 , Number 1 , 1 993

19

Fig. 1. Paired leaf-pouch domatia, covered with dense, erect trichomes, at the base of a leaf of Delpydora
macrophylla Pierre (Sapotaceae) in southern Cameroon. These pouches are formed by downward folding and
rolling of the expanded base of the blade on either side of the midrib. The domatia are usually occupied by timid
Technomyrmex species.

Stipules have been modified into ant-domatia in
a few myrmecophytes; known examples are all
from the Old World tropics (Appendix 1 ). (The
only apparent exception is Acacia, in which thorns,
themselves highly specialized stipules, have been
modified into domatia in both neotropical and Old-
World representatives.) In many tropical plants,
large stipules function as mechanical protection for
the growing bud. In some cases, stipules possess
ant-attractive structures which provide biotic de-
fense as well. Where stipules are persistent, rather
than being shed soon after maturation of associated
nodes, ants may find suitable shelter for tending
homopterans, nesting, or both. Although ants and
their associated debris are observed frequently be-
neath large stipules, only rarely have these stipules
become evolutionarily modified to house ants.
Specializations include recurving or inflating of the

stipule to form a more enclosed structure (as in New
Guinea Psychotrio and perhaps African Docty-
Jodenio), location of specialized food bodies on the
lower surface of the stipule (Asian Mocaranga),
and possibly the evolution of persistent stipules. In
an analogous case, African Diospyros conocorpo
Giirke & K. Schumann has specialized, hairy
domatia formed from persistent cataphylls
(Letouzey and White 1970). These structures are
leaf-like appendages, usually rapidly deciduous,
and formed on the first few nodes of young expand-
ing twigs in many tropical trees with rhythmic
growth patterns (Halle et al. 1978). They are func-
tionally analogous to stipules. In D. conocorpo , the
cataphylls are folded to form a structure completely
enclosed, except for a small opening near the base
of the blade, and they are persistent, rather than
deciduous, as in related species. These structures

20

Journal of Hymenoptera Research

are occupied by Technomyrmexkohlii ( Forel), which
also inhabits several leaf-pouch ant-plants in the
same forests.

In Asia, Africa, and the Neotropics, leaf-pouch
domatia of strikingly similar form have evolved in
numerous myrmecophyte lineages (Appendix 1 ).
Formed near the leaf base, and typically paired on
either side of the midrib (but single in some spe-
cies), they are usually covered with long, dense
trichomes (Table 1 ). Restricted to understory treelets
and shrubs, these ant-plants typically are occupied
by small, timid ants. Leaf-pouches seem to be
formed in one of two ways. In some taxa (e.g.,
neotropical Melastomataceae, and African
Sterculiaceae), invagination occurs in the internal
portion of the leaf blade, in a region flanking the
base of the midrib. This invagination produces
single or paired inflated pouches, each with an
entrance on the abaxial leaf surface. In at least four
plant families, including most frequently and vari-
ably in the Rubiaceae, paired leaf pouches form in
a different manner. At the bases of leaf blades,
(revolute) leaf margins curl downward, as in Afri-
can Delpydora (Fig. 1), Magnistipula, Dichape-
tahwi gassitae Bret., and I.xora hippoporifera
Bremek., neotropical Hirtella and Remijia , and
Asian Callicarpa saccatci Steen. Less frequently,
(involute) leaf margins curl upward, as in neotropical
Duroia sacciferci Benth. and Hook. Pouches may
be bubble-like invaginations (Gardenia imperialis
L. Pauwels) or, more often, scroll-like hollow tubes.

It has long been postulated that the leaf-pouch
domatia of ant-plants evolved from acarodomatia
(Schnell 1966, Schnell et al. 1968), presumably by
intermediate stages in which domatia could be
occupied either by mites or by small ants. Selection
led to increased size of domatia with progressive
transference of protective function from mites to
ants (O’Dowd and Willson 1989). Benson (1985)
also argues that leaf-pouch domatia evolved in
myrmecophytes from small depressions in leaf
surfaces. The original function of these depres-
sions was to shelter ant-tended homopterans. The
two hypotheses are not mutually exclusive, as ants
may also have used acarodomatia to shelter ho-
mopterans (Benson 1985). Hypotheses implicat-
ing acarodomatia in the origin of leaf-pouch ant-
domatia receive strong support from cases like

Cola tnarsupiwn K. Schumann, in which a single
leaf presents a graded series of domatia increasing
in size from typical acarodomatia at the leaf apex to
large inflated pouches at the leaf base (Schnell and
Beaufort 1966).

Why have leaf-pouch domatia evolved repeat-
edly in certain groups, for example, at least nine
times in the tribe Miconieae in the Melastomataceae
(Table I)? Leaves of many Miconieae have strongly
arcuate venation with sections of the leaf blade
vaulted and curved upward between major veins.
Even before selection intervened to enlarge these
structures, this waffle-like leaf organization may
have fortuitously provided invaginations large
enough to shelter ant nests. In African Sterculiaceae,
where similar domatia have evolved twice, vena-
tion is also palmate, with three large veins converg-
ing at the leaf base.

The largest group of myrmecophytes is that in
which domatia are located in stems, or in stem-like
structures such as petioles or inflorescence stalks
(Appendix 1). Increasing evidence supports the
hypothesis that ants originally colonized cavities
created in twigs and petioles by wood-boring in-
sects (Ward 1 99 1 , also Appendix 1 ). Together with
cavities formed by spontaneous drying of pith ca-
nals, these cavities provided ants with shelter and
substrate for brood and symbiotic Coccoidea. When
the presence of ants conferred net benefit (e.g., by
protection against phytophagous insects, including
wood-borers, and any diseases transmitted by these
insects), selection acted on the plant to evolve
features facilitating its occupancy by ants (Ward
1991). Such traits include specialized swollen
twigs and a prostoma, or relatively unlignified spot
through which ants gain easy access to the domatia.

What traits may have predisposed plants to
evolve symbiotic association with ants via this
mechanism? Wood-boring insects usually attack
soft, pithy portions of stems. The larger the primary
diameter of a stem, the thicker its pithy central
section. Thus thick-twigged plants offer greater
opportunities than do thin-twigged taxa for wood-
boring insects, and for ants which nest secondarily
or primarily in the cavities of living plants. Al-
though much poorly understood interspecific varia-
tion in stem structure affects the relationship be-

Volume 2 , Number 1 , 1993

21

tween the primary diameters and pith diameters of
twigs, myrmecophytes are most likely to evolve in
plants with thick twigs.

This observation gains importance when we
consider the plant-architectural correlates of stem
primary diameter. The best known of “Corner’s
rules,” and one confirmed by quantitative studies
(White 1 983 ), states that there is a positive correla-
tion between the primary diameter of a stem axis
and the size of appendages (e.g., leaves) borne by it
(Halle et al. 1978). This correlation means that
selection acting on leaf size (Givnish 1987) also
drives evolutionary change in stem diameter
( McKey 1991). Thus, the evolution of stem domatia
may be facilitated by an evolutionary increase in
leaf size, driven for example, by climatic change,
by range extension into more mesic environments
(Givnish 1987), or by selection to minimize meta-
bolic cost of woody leaf-support tissues (White
1983). If disparities in leaf size were related to
habitat, myrmecophyte frequencies could be corre-
lated with habitat, independently of and perhaps
even despite any habitat-related differences in se-
lection imposed by symbiotic ants (McKey, unpub-
lished).

Corner’s Rule may help account for several
groups of ant-plants with domatia in thickened
support structures (Appendix 1 ). First, myrmeco-
phytism has evolved often in genera whose moist,
shaded, understory environments have favored com-
paratively large, broad leaves and thick stems (e.g.,
African Leonardoxa, and Oriental or Australian
Tapeinosperma, Steganthera , Kibara, and
Myristicci). Ants also live symbiotically with mem-
bers of the Meliaceae, Sapindaceae, and
Anacardiaceae, whose leaves are not only large, but
compound. In the Meliaceae, myrmecophytes ap-
pear to have evolved independently in four genera,
including three Asian taxa ( Aphanamixis ,
Chisocheton and Aglaia) with massive stems sup-
porting large compound leaves. Even within
Aphanamixis, myrmecophily characterizes forms
with relatively large leaves and twigs (Mabberley
1985). Second, thick support structures for large
leaves may also have facilitated the frequent evolu-
tion of ant-plants in fast-growing pioneer trees,
whose large leaves and sparse branching allow
them to support a considerable leaf surface area

with minimum investment in woody framework
(White 1983). Examples are neotropical Cecropia,
Asian Macanmga and Australian Endospermum ,
which almost surely converged due to selection on
leaf size and tree architecture prior to the evolution
of myrmecophytism. Other myrmecophytic pio-
neers of riverine and forest light gaps include
neotropical Triplaris, Australian Nauclea and Afri-
can Barteria and Vitex grandifolia Giirke. In all of
the plants in these two categories, ant protection
might be especially advantageous, because the large
and parenchyma-rich meristems are especially sus-
ceptible to damage by wood-boring insects. Since
most of these plants produce one-to-few large mer-
istems at any one time, the material and opportunity
costs of losing even one meristem could be very
high.

Finally, two smaller groups of ant-plants house
ants in either false nodes, thickened to support
multiple leaves (e.g., two Cordia species and Diiroia
hirsuta Poepp. and Endl.), or in stout petioles
{Piper, Pourouma and Tachigali). Although peti-
oles might often be too short-lived to function as
domatia, they are likely to be comparatively long-
lived for both the compound leaves of Tachigali
and the simple leaves of myrmecophytic under-
story Piper species (in which ant cavities also
extend into the stem itself).

Preadaptations and Pathways to
Specialization in Ants

Specialized plant-ants are represented dispro-
portionately in particular taxonomic categories of
ants, and shared characteristics of these taxa pro-
vide evidence of factors predisposing ants to evolve
symbiotic relationships with plants. Worldwide,
plant-ants have evolved in five of 1 2 subfamilies in
the Formicidae (Appendix 1 ). They are absent only
from subfamilies of specialized legionary and other
predatory ants (Cerapachyinae, Dorylinae,
Ecitoninae, Leptanillinae, and Myrmeciinae), and
from the monotypic Aneuretinae and Nothomy-
rmeciinae. Until recently, they were also deemed
absent from the Ponerinae, the most predatory of
five subfamilies containing at least some species
that depend directly and substantially on plant
resources. However, at least four species of

22

Journal of Hymenoptera Research

Fig. 2. Leaves bound together with carton to form the ephemeral nests of Dolichoderus (= Hypoclinea) bidens (L.)
in southeastern Peru.

Pachycondyla now appear to be specialized sym-
bionts of Cecropia ( Davidson et al . 1991, Davidson
and Fisher 1 99 1 , J. Longino, personal communica-
tion). Still, plant-ants are poorly represented in the
Ponerinae and among predatory ants in general.

The evolution of obligate plant-ants in five sub-
families, approximately 30 genera (Appendix 1 ),
and multiple clades of at least Psendomyrmex (Ward
1991) and Azteca (Benson 1985, Longino 1991a
and b) confirms the frequency and facility with
which plant-ants have evolved, and provides abun-
dant opportunity to find commonalities in lifestyles
and traits that may have promoted evolutionary
specialization on plants. For example, three of the
six principal generic radiations of South American
endemics have arisen (one each) in the sub-family
Pseudomyrmecinae, and in the tribes Tapinomini
(Dolichoderinae) and Myrmelachistini (Formi-
cinae). These ants share the habit of regularly
tending homopterans inside (all three taxa) or out-
side (especially tapinomines) of cavities in live
plants. Within each of these groups, common
ancestors of contemporary plant-ants likely had
additional traits which predisposed them to evolve
symbiotic (parasitic as well as mutualistic) associa-

tions with homoptera and plants. Because the
relative competitive abilities of ants form an impor-
tant part of the story, we turn now to consider
various ecological differences among ants with
different competitive abilities.

Competitive Dominants . — Ecological limita-
tions on populations of arboreal ants in lowland
tropical forests add insight into probable origins,
correlates and consequences of arboreal nesting
habits, including stem-nesting. Colony popula-
tions appear to be limited principally by food and
nest sites (Wilson 1959b, Carroll 1979, Davidson
and Epstein 1989). Because most arboreal ants are
generalized foragers of plant and homopteran exu-
dates, and of carrion, interspecific food require-
ments are strongly overlapping, and competition
can be intense. The competitive dominants of each
tropical biogeographic region are species which
have evolved means of nesting in areas of abundant
food. They include Old World Oecophylla,
Crematogaster, Tetramorium, Philidris and
Polyrachis, some Austral ian Anonychomyrma, and
New World Crematogaster, Camponotns, Azteca
and Dolichoderus (including Hypoclinea , Shattuck,
1 992a). These ants either bind leaves together into

Volume 2, Number 1, 1993

23

temporary nests, or construct potentially more per-
manent carton homes in the canopy where food is
abundant (Fig. 2). Like species which occupy the
top of the competitive hierarchy at high temperate
latitudes (Vepsalainen and Pisarski 1982), these
species defend not only their nest sites and tempo-
rary, localized food patches, but their entire forag-
ing areas, as absolute territories. Although a certain
threshold of aggressiveness may have been re-
quired before these ants could defend their some-
what exposed nests successfully against vertebrate
enemies (e.g., monkeys and woodpeckers; J.
Longino, personal communication), an eventual
capacity to nest near abundant food almost cer-
tainly contributed to the escalation of aggressive-
ness and dominance.

Most competitive dominants tend populations
of Homoptera, whose exudates form a steady and
predictable source of colony nutrition and help to
fund high worker activity and aggression. These
ants lack functional stings, but all possess elaborate
chemical weaponry (Blum and Hermann 1978,
Attygalle and Morgan 1984, Buschinger and
Maschwitz 1984, Merlin et al. 1992). Expended in
use, these exocrine products should be character-
ized by more rapid turnover and greater cost than is
associated with longer-lived stings and sturdy ex-
oskeletons. Nevertheless, if chemical defenses are
supported by the requisite resource base, they ap-
pear to be more effective than stings in contests
among ants (Davidson et al. 1988). With their rich
sources of homopteran exudates, dominants should
often experience an excess of dietary carbon in
relation to protein, so that colony expansion is
protein-limited. If so, this could explain the “high
tempo” lifestyle (sensu Oster and Wilson 1978)
characteristic of these ants and help to resolve the
enigma of their seeming “inefficiency” in foraging
(Oster and Wilson 1978; Holldobler and Wilson
1990). By spending relatively “cheap” carbon
resources on aggression and seemingly extravagant
levels of activity, these ants secure dominance over
territories whose protein resources fund colony
growth.

Chemical weaponry and high activity levels are
not the only traits determining dominance in these
ants. Abundant food and freedom from nest site
limitation appear also to have led to larger colony

sizes and longer life expectancies. If the resource
environments of ants have helped to shape the
evolution of life history attributes (e.g., rates of
egg-laying, worker turnover, etc.), a correlated
evolved dependency on rapid rates of resource
acquisition may restrict some dominants to the
most productive sites in lowland rain forests
(Davidson and Epstein 1989). Arboreal dominants
are preeminent in monopolizing high quality re-
sources at exposed sites such as EFN’s, and
Homoptera positioned on flowering and fruiting
peduncles, where a plant's phloem resources are
frequently most concentrated. As evidence of their
competitive impact in one rainforest ant commu-
nity as a whole, Wilson (1959b) noted that a num-
ber of arboreal ant species regularly forage on the
ground, whereas only a few exceptional ground
nesters forage even in the low arboreal zone, and
possibly none of these reaches the upper canopy.

( In the Neotropics, terrestrially nesting Paraponera
and Ectatomma are obvious counter-examples, but
both genera are exceptional among ponerines for
their heavy reliance upon plant exudates, carried as
large droplets in the mandibles.) Dominants can
restrict the local diversity of other ants, as do the
parabiotic associates of neotropical ant-gardens
(Davidson 1988). Thus, the diversity of arboreal
but not terrestrial ant species is lower inside the
territories of Ccnnponotus femoratus Fab. and
Crematogasterc f. linmta parabiotica (Forel), than
in adjacent areas lacking these ants. Because the
species composition and diversity of subordinate
species often varies markedly with the identity of
dominants, patchiness in the territories of domi-
nants determines a mosaic of ant communities
within many tropical forests (Leston 1973, re-
viewed in Holldobler and Wilson 1990).

Competitive dominance may be context depen-
dent, e.g., differing in relation to the identities of
plant species which form the substrate for ant
nesting and foraging (Davidson and Epstein 1 989).
Thus, host plant associations of Oecophylla
longinoda (Latr.) (Dejean and Dijicto 1990) and
Tetramorimn aculeatum (Mayr) (Dejean et al. 1 992),
two widespread dominants in African forests, are
correlated with worker preferences for foliage types
offered in laboratory experiments.

24

Journal of Hymenoptera Research

Weak Competitors. — For competitively subor-
dinate ants, the benefits of combining nesting and
foraging locations are conditional on locating nests
and resources in sites which are protected from
invasion by dominants. Nests in dead or live twigs,
stems, and larval insect borings can be defended if
the cavity size is not much larger than the head
diameters of workers, soldiers, or queens. By
sealing a stem nest with her head, a single worker
can protect her whole colony or colony fragment
from invasion by enemy ants. Thus, along a Pacific
Ocean beach at Corcovado National Park in Costa
Rica, the many dead twigs of Coccoloba
(Polygonaceae) trees were occupied by more than
nine ant species (six of Pseudomyrmex alone),
whose head widths were roughly equal and propor-
tional to the internal diameters of their twig cavities
(D. Davidson, personal observation). At least some
African Tetraponera also appear not to nest in
stems whose diameter exceeds a threshold value
(Terron 1970). For small-bodied ants like the timid
Wasmannia scrobifera Kempf in Costa Rica, other
protected sites may include carton shelters beneath
leaves of plants whose dense stem trichomes ex-
clude larger bodied workers (see below).

For comparatively docile and subordinate ants,
the advantage of locating their resources inside
stem cavities is clear. The evolutionary transition
from nesting in dead twigs to nesting in live twigs
and other cavities of live plants conveyed the addi-
tional opportunity to obtain uncontested resources
from phloem-feeding Homoptera (especially
Coccoidea), which either invaded such cavities on
their own or were brought there by the ants. More-
over, nests in live wood were potentially habitable
over much longer time periods than those in dead or
decaying twigs and branches, obviating a need for
frequent and dangerous nest moves. Longer tenure
of living nest sites, which grew rather than decay-
ing with time, may secondarily have allowed the
evolution of larger colony sizes and increased op-
portunities for local monopolization of resources,
as well as the selective advantage of aggressive
behavior and allelochemicals. Traits conferring a
capacity to nest in live plants are not well studied,
but they probably involve evolutionary adjustment
to an increased threat from nest pathogens. Thus,
Ward (1991 ) points out the tendency for hypertro-

phy of metapleural glands in domatia-inhabiting
pseudomyrmecines. Where studied, the function
of metapleural secretions has been tied to the sup-
pression of microbial pathogens (e.g., Maschwitz
1974, Holldobler and Engel-Siegel 1984).

In summary, plant-ants are most frequent in taxa
which depend directly or indirectly, but substan-
tially, on plant resources. They are most likely to
have evolved in competitively subordinate ants,
selected to live in close proximity to food re-
sources, but to nest and feed in comparatively
protected and permanent sites which reduce dan-
gerous contact with competitive dominants. Within
this subset of ant taxa, selection for evolutionary
specialization of plant-ants might have been less
likely in groups where potent defensive exocrine
compounds (e.g., many Dolichoderus species), and
worker armor or specialized diets (e.g.,cephalotines)
diminished the hazards of encounters with domi-
nants.

The transition from more generalized ancestors
to specialized plant-ants would not have been dif-
ficult. Founding queens should have evolved greater
efficiencies in locating hosts that provided superior
food or housing or were more easily accessed. By
consuming or deterring insect herbivores, ants might
then have enhanced their own fitness indirectly by
promoting the vigor or prolonging the lifespans of
their hosts. However, host specialization by ants,
as well as consumption of eggs and larvae of insect
herbivores, could have been favored in ants whether
the ants and Homoptera had a net positive or nega-
tive effect on these hosts. Longino (1987), for
example, discusses the case of Leptothorax obtura-
tor Wheeler, which nests only in cynipid galls of
oaks and probably has no fitness effect on the host
tree. Selection pressures on ants and plants should
often have been asymmetric, leading to the expec-
tation that ant-attractive traits would have evolved
in only a subset of the host plants on which obligate
plant-ants reside.

Expectations based on this brief review of com-
petitive interactions among arboreal ant species can
now be compared with actual patterns in the distri-
bution and ecology of specialized plant-ants.

Volume 2, Number 1 , 1 993

25

THE MATCHING OF ANTS AND PLANTS

Appendix 1 is a worldwide summary of all
symbiotic ant-plant relationships known to us. To
facilitate comparisons among ants of differing
lifestyles and competitive abilities (see below), we
organize the data by ant genus. Also evident in this
summary is the basic asymmetry in the degree to
which relationships are obligate for ants versus
plants. The vast majority of the ants in the table are
thought to be obligate plant-ants (column 7, though
these are not necessarily host-specific). However,
a substantial fraction of their host genera have no
obvious myrmecophytic traits (column 6), despite
their regular association with specialized (usually)
or unspecialized ants (cf., African Musanga and
Neotropical Tetrathylaciwn). Few plants with con-
spicuous myrmecophytic traits (e.g., obvious
domatia, or naturally hollow stems with prostomas)
lack specialized plant-ants altogether, though some
may occur principally with unspecialized ants in
marginal habitats, or at the edges of their distribu-
tions (see below).

Almost certainly. Appendix 1 includes a mix of
relationships in which ants are parasitic,
commensalistic, or mutualistic with their hosts, and
the net outcome of the interactions might even vary
with habitat or ecological context. These outcomes
are not wholly predictable from myrmecophytic
traits, since even in mutualistic associations, plants
need have no obvious specializations to attract ants.
Clearly most of the relationships are poorly known,
and many of the table entries are incomplete. Yet
the table clarifies the types of data which will
eventually be essential to describe pattern in these
relationships, and we hope it will stimulate the
collection of such data in future studies.

Despite limited data, patterns in relationships of
ants and host plants correspond roughly to those
noted for tropical forest ant faunas as a whole.
Across genera, the fastest growing myrmecophytes
of disturbed forest edge (i.e., hosts with rapid rates
of resource supply to ants) tend to be inhabited by
ants from aggressive, dominant, carton-building
genera (column 1), e.g., the Azteca of neotropical
Cecropia, and Crenmtogaster of ecologically simi-
lar Old World (especially Asian) Mcicaranga. Less
aggressive and competitively subordinate ant spe-

cies tend to persist by employing one or more of
several strategies likely to reduce interactions with
the dominants. We deal with each of these in turn.

Ant Pruning of Host-plant Neighbors

The most common and significant natural en-
emies of ants are other ants (Haskins 1 939). Species
with sting defenses, usually inferior to chemical
defenses in contests among ants, are disproportion-
ately likely to attack and prune vegetation sur-
rounding their hosts (Davidson et al. 1988, also
column 4, Appendix I). In both Africa and South
America, this behavior is most widespread in
pseudomyrmecine plant-ants, where pruning has
evolved multiple times in independent lineages
(Ward 1990). The potent stings of
pseudomyrmecines may be an effective deterrent
of vertebrates ( Janzen 1 972 ), but they are inferior to
chemical defenses in repelling colonies of invading
ants. Although the Pseudomyrmex of Triplaris and
Acacia , and the Tetraponera of Bacteria, do not
forage extensively off their host plants, they regu-
larly leave these hosts to sever the petioles of leaves
on neighboring plants (Fig. 3). Eventually these
neighbors die, leaving the host trees in starkly
defined clearings within the forest.

Such clearings have been hypothesized to re-
ward resident colonies by enhancing host-plant
vigor or, in drier environments, acting as natural
fire breaks (Janzen 1967a). However, experimen-
tal evidence suggests that a more immediate selec-
tive advantage for attacks on neighboring vegeta-
tion is the reduction of threats from more dominant
arboreal ants. When permanent wire bridges were
made between myrmecophytic Triplaris and neigh-
boring trees, the frequency of invasions by domi-
nant Crenmtogaster increased, and whole hosts or
portions of these hosts were eventually usurped by
Crenmtogaster or Azteca species (Davidson et al.
1988). The broad taxonomic distribution of obli-
gate and facultative pruning behavior (the latter
occurring only in the presence of enemy ants.
Appendix 1, column 4) suggests that dominant
competing and predatory ants constitute a major
threat to many or most specialized plant-ants. Its
prominence in neotropical ants is evidence against
the hypothesis that a paucity of dominants charac-

26

Journal of Hymenoptera Research

terizes that region (Carroll 1979, see also Me Key
and Davidson, in press). Presumably, pruning be-
havior could also serve to defend resident colonies
against invasions by leafeutter (Morawetz et al.
1 992) and legionary ants, which could devastate the
resource base or the colony itself.

Pruning behavior is not strictly limited to ants
with functional stings (Appendix 1). Most
neotropical Cecropia and Old World Macaranga
and Endospemntm establish in disturbed second
growth vegetation, where vines are particularly
abundant and troublesome to both plants and ants,
and where weedy dominant ants are a constant
threat (Benson 1985). Not surprisingly, the com-
mon ant associates of these host genera ( Azteca ,
Crematogaster and Camponotus, respectively) will
attack encircling vines (Appendix 1, Janzen 1969,
Fiala et al. 1989; Davidson personal observation,
Letourneau et al. 1993), though pruning is not
typical for these genera as a whole. The compara-
tively unbranched growth forms of these hosts may
also help to limit contact with vines and neighbor-
ing plants (Putz and Holbrook 1988) and, therefore,
with enemy ants (Benson 1985). In contrast to
chemically defended ants, in which pruning is
restricted to species inhabiting hosts of secondary
forest, species defended principally by strong or
weak stings (Pachycondyla, Tetraponera,
Pseudomynnex, and Allomerus) also tend to prune
around hosts in primary forests, where threats from
vines and dominant ants are not so severe. Not all
plant-ants in these genera prune, but some species
benefit from other forms of protection (see below).

Worldwide, the most dramatic case of allelopa-
thy by ants may be that of Mynnelachista
(Formicinae) species inhabiting myrmecophytes in
the intriguing western Amazonian “Supay chacras”
(Quechua for “Gardens of the Devil”). Dominance
of lowland forest stands (to > 1 0,000 nr in size ) by
multiple species of myrmecophytes, most promi-
nently Duroia hirsutci [Poeppig and Endl.] K.
Schunr, but also Cordio nodosa Lam. and Miconia
nen’osa Triana, suggests that the ants kill non-
myrmecophytes selectively (Campbell et al. 1989).
In a similar phenomenon, at somewhat higher el-
evations of western Amazonia (700-1200 m), a
different Mynnelachista species creates monospe-
cific stands of myrmecophytic Tococa occulentalis

Naudin (Morawetz et al. 1992). The two conge-
neric ants share a similar behavioral ecology (D.
Davidson, personal observation, for supay chacras,
and Morawetz etal. 1992, for Tococa). Workersdo
not appear to forage off their hosts, but do leave
their hosts to attack other plants. When seedlings or
saplings of plants other than the host species are
placed in the vicinities of these hosts, workers gnaw
at the vascular bundles of leaves of the introduced
plants, and can kill them in a matter of hours to days.
Morawetz and colleagues describe the extraordi-
nary capacity of these ants to single out especially
vulnerable plant tissues for attack. Thus, workers
bite and poison palmate leaves at the base of lami-
nae, where all vascular bundles join, pinnately
nerved leaves at nerve bases of the first and second
order, and monocots (e.g., palms), nerve by nerve,
along the entire leaf. Necrosis originating at the
attack sites spreads rapidly over the entire lamina.
Within a few hours to a few days, inhabitants of the
Tococa can successfully kill seedlings and saplings
within a radius of 4 m and damage trees up to 10 m
in size. Light gaps created by ant activities are
subsequently colonized by vegetative propagation
of the host.

Although Morawetz and colleagues discount
the hypothesis that the killing of host plant neigh-
bors by Mynnelachista has evolved principally to
exclude enemy ants, several observations suggest
that the hypothesis should not be ruled out. First,
leaf-cutter ants, an important enemy of the Tococa ,
invade principally by contact with the branches of
other plants, not via the main trunk. Second, no
generalized arboricolous ants appear to forage within
the territories of these specialized Mynnelachista.
Furthermore, large worker forces may be needed to
assure the safety of ants which have left their hosts.
Attacks on neighbors of Tococa begin when the ant
population of one or a few individual hosts is at
least 1500 workers in size. Similarly within supay
chacras, smaller fragments of the extended colo-
nies show extreme fidelity to their individual hosts,
and only workers of the largest trees leave their
hosts to swarm over seedlings and other vegetation.
Moreover, the latter activities appear to be re-
stricted to hot and sunny conditions (D. Davidson,
personal observation), which may allow maximum
worker activity and performance levels. To date.

Volume 2 , Number 1 , 1 993

27

Fig. 3. Pseudomyrmex dendroicus Forel on branches of neighboring plants, whose leaves have been pruned by the
ants. The long, thin body shape of workers in Pseudomyrmex spp. may preclude their use of plants with long, dense
trichomes.

there have been no experimental tests of the effects
of creating artificial and unseverable bridges be-
tween neighboring intact trees and hosts of these
Myrmelachista. Such experiments would greatly
aid in assessing the evolutionary significance of the
extraordinary behavior of these ants.

Hiding among Trichomes

The long, dense and erect trichomes on stems
and domatia of many myrmecophytic plants form
mechanical barriers to the movements of large-
bodied ants and create safe havens for colonies of
obligate plant-ants with timid and diminutive work-
ers (Davidson et al. 1 989; Fig. 4, Appendix 1 ). Ant-
plants with inhibitory hairs on stems, domatia or
both, occur in at least 18 neotropical genera (eight
within Melastomataceae alone), and eight families,
and appear to have evolved independently on at
least 21 separate occasions (Table 1). In Africa,
such hairy ant-plants occur in at least eight genera
and six families, with each generic occurrence
representing a single independent origin. Trichome-

myrmecophytes have also evolved in at least four
genera in the Oriental and Australian tropics (Table
1 ), though the symbiotic associates of these plants
remain unknown. In many or most genera of hairy
ant-plants, long, erect pubescence also occurs in
non-myrmecophytic congeners. It therefore seems
likely that docile, small-bodied ants initially sought
safe nesting and foraging sites on hairy plants prior
to the evolution of myrmecophytism in these lin-
eages. A possible contemporary example of such a
relationship is that between Wasmannia scrobifera
and a non-myrmecophytic hairy Piper species in
Costa Rica (D. Davidson, personal observation).
These ants build small fragile carton nests on abaxial
leaf surfaces, where they feed on pearl bodies.
Nests are not limited to individual host plants, nor
are the ants likely to be obligate plant-ants. In some
cases, ant dependency on plant trichomes may be
restricted to the early stages of colony foundation.
Thus, certain Azteca species regularly initiate colo-
nies on pubescent ant-plants like Cordia and Tococa
but later prune runways through host-plant tri-
chomes and form carton satellite nests on neighbor-
ing trees lacking protective trichomes (“i” in col-

28

Journal of Hymenoptera Research

Fig. 4. Tiny Pheidole minutula Mayr workers travel easily among the erect trichomes of this myrmecophytic
Clidemia. Numerous ant species with tiny workers use such “trichome myrmecophytes” as protected feeding and
nesting sites, where they are safe from larger-bodied competitors and predators.

umn 4 of Appendix 1 ; D. Davidson, personal obser-
vation, Benson 1985).

Contemporary distributions of ants across
myrmecophytes in Africa and the Neotropics illus-
trate the influence of plant trichomes on the match
between ants and plants (Appendix 1). First, in both
regions, worker ants of pubescent myrmecophytes
are short-bodied (<3 mm), with short turning radii,
and do not include longer-bodied pseudomyr-
mecines. Included here are two neotropical genera
with functional stings ( Allomenis and Solenopsis),
and docile African dolichoderines in the genus
Technomyrmex (species formerly placed in
Eii gramma, Shattuck, 1992a). All known hosts of
Allomenis and Solenopsis possess long erect pu-
bescence. Allomenis is particularly conspicuous in
its association with a diversity of pubescent host
genera, seven in total. Of the recorded hosts of
African Technomyrmex , species in five (and possi-
bly six) of eight genera are hairy; only two,
Leonardoxa and Ixora hippoporifera, definitely
lack trichomes. To the extent that members of
competitively dominant ant genera depend on pu-

bescent ant-plants beyond the incipient colony stage,
the particular species represented in these associa-
tions are unusually timid for their genera (e.g., the
Crematogaster cf. victima group on melastomes,
the tiny Crematogaster sp. on Delpydora , and the
Azteca species inhabiting hairy Triplaris
poeppigiana Weddell). Second, the body sizes of
plant-ants tend to be correlated with trichome spac-
ing (Davidson et al. 1989). This suggests that
ancestral ants may have nested preferentially not
only on pubescent plants but specifically on those
where mean distances between trichomes were no
larger than required by their own body sizes. The
parallels with nest selection by stem diameter in
generalized stem-nesting ants are obvious (see
above).

Third, if ants compete for host plants (see
Davidson et al. 1989), and if small, timid species
persist only where protected by trichomes from
larger dominants (> 3 mm, e.g., Crematogaster and
Azteca ), then the dominants should prevail on
myrmecophytes lacking inhibitory trichomes. This
hypothesis is supported not only across ant-plant

Volume 2, Number 1, 1993

29

genera (Appendix 1), but within several genera
which are interspecifically variable in pubescence.
In neotropical Cordia , for example, glabrous C.
alHodora (R. and P.) Oken is regularly occupied by
aggressive Azteca, but smaller and more timid
Allomerus ants inhabit densely hairy C. nodosa. As
noted above, a small-bodied and timid Azteca spe-
cies inhabits the hirsute stems of Triplaris
poeppigiana, though the vast majority of
myrmecophytic Triplaris species are glabrous and
occupied by long and narrow-bodied pseudomyr-
mecines. Third, dominant Crematogaster ants oc-
cupy glabrous African Canthium, whereas hairy
congeneric hosts are associated with timid
Teclinomyrmex species (Bequaert 1922, pp. 474-
475). The same may perhaps be true in African
Cuviera , which contains both glabrous and hirsute
myrmecophytes. Both Teclinomyrmex and
Crematogaster are recorded as associates of ant-
plants in this genus, but the distribution of different
ants in relation to plant pubescence cannot be
discerned from existing literature. Finally, as noted
above, some ants 3 mm in body length occasion-
ally occupy trichome myrmecophytes but regularly
prune trail systems, which facilitate their move-
ments (Davidson et al. 1989).

Association of Camponotus ants with spiny
palms in the genus Korthalsia may also have had its
origins in the tendency of ants to feed and nest
where the plant's growing tips are protected from
the ants’ natural enemies. Among the 12 Korthalsia
species which Dransfield (1984) lists for Sabah,
Malaysia, seven have armed ocrea and five do not.
Of the species with spiny ocrea, all but K. ferox
Becc. also show regular associations with ants,
whereas this is true for none of the species with
unarmed ocrea. Both the long, sharp and compara-
tively dense spines of species K. echinometra Becc.,
K. hispida Becc. and K. robusta Blume, and the
scattered, short, triangular spines of K. cheb Becc.,
K. furtadoana J. Dransf. and K. rostrata Blume are
more likely to protect the ants from vertebrate
predators than from other ants. Dransfield (1981)
found greater herbivory by vertebrates (perhaps
squirrels) on growing tips of K. rigida Blume (with
unarmed ocrea and sparsely armed leaf sheaths)
than on those of K. echinometra and K. rostrata.
Although he attributed this result to protection that

ants might afford to the latter species, an alternative
hypothesis is that both the plants’ growing tips and
the ant nests benefit from the armature of ocrea and
leaf sheaths. This would not rule out some addi-
tional benefit to the plant from its ants. Unfortu-
nately, phylogenetic relationships remain unde-
fined for both plants and ants, and it is not yet
possible to determine the extent to which the vari-
ous relationships between ants and armed Korthalsia
evolved independently. However, Dransfield’s
(1981) observation that Calamus species of New
Guinea and the Philippines show parallel evolution
of armed ocrea and relationships with ants suggests
that myrmecophytism could have evolved more
than once within Korthalsia as well. Similarly,
myrmecophytic rattans in the genera Calamus and
Daenionorops exhibit parallel evolution of ant gal-
leries formed by interlocking combs of spines,
forming collars on the leaf sheaths (Dransfield and
Manokaran 1978).

Rates of Resource Supply from Plants

The impact of rates of resource supply on the
match between ants and plants is best compared
within host genera, holding food type approxi-
mately constant. Within western Amazonia, for
example, the rate of food body production by Ce-
cropia varies across both species and habitat types
(Davidson et al. 1991, Davidson and Fisher 1991,
Folgarait and Davidson 1992). Faced with compe-
tition from fast-growing pioneer species of similar
stature, more light-demanding species of large riv-
erine disturbances defer costly defense in favor of
rapid growth. Because comparatively shade-toler-
ant species of small forest light gaps experience
light competition from much larger neighbors, di-
version of limiting carbon from defense to growth
might confer little benefit, and even jeopardize the
persistence required to take advantage of later
canopy openings. Thus, the more shade-tolerant
Cecropia species produce swollen stems, prostomas,
and trichilia much earlier in development than do
their light-demanding close relatives (Fig. 5), as
well as producing a greater dry weight of Mullerian
bodies per unit leaf area. Despite this greater
investment (proportional to the plant’s resource
budget) in biotic defenses by small gap Cecropia,

30

Journal of Hymenoptera Research

there are at least three reasons why the absolute
rates of food provisioning to ants are greater in
light-demanding pioneers than in closely related
but more shade-tolerant gap species. First, and
perhaps foremost, the smaller sizes of forest gap
species at the time of colonization by ants are
associated with fewer leaves (sources of food re-
wards ) and slower pi ant growth rates. Second, even
with plant size or light environment held constant,
small gap Cecropia have intrinsically slower growth
and leaf production rates than do their more light-
demanding counterparts. Finally, comparatively
low light intensities in their typical habitats further
limit the capacity of the forest gap plants to produce
ant rewards.

Ants appear to respond to these quantitative
differences in food production rates of Cecropia
(Davidson et al. 1991, Davidson and Fisher 1991).
For example, in southeastern Pern, patterns of ant
associations are more closely tied to habitats than to
host identities. Although the closest taxonomic
relationships appear to be between Cecropia in
different habitats (C. C. Berg, personal communi-
cation), Aztecao\'aticeps¥oxc\ inhabits only intrin-
sically fast-growing pioneers of riverine and stream-
side habitats. In contrast, specialized Camponotus,
Pacliycondyla and Crematogaster species, and
Azteca australis Wheeler are the typical residents
of relatively slow-growing and congeneric hosts of
small light gaps. Although the latter ants frequently
colonize riverine Cecropia , they seldom establish
colonies there, and they may usually be outcompeted
by rapidly developing colonies of A. ovaticeps.
This pattern holds both within and across host
species, and it suggests that ant species may coexist
locally by virtue of their “included niches”. Spe-
cies with rapidly growing colonies may dominate
higher quality hosts, but be unable to tolerate low
rates of resource supply. On the other hand, ants
with relatively slow-growing colonies tolerate both
high and low quality resources, but are usually
excluded by competitors from fast-growing hosts.

A similar pattern of niche differentiation is ap-
parent within plant-ant guilds of other
myrmecophyte taxa (including epiphytes) of both
the New and Old World (Davidson and Epstein
1989, Davidson et al. 1989 and 1991). For ex-
ample, specialized Tetraponera are the typical resi-

dents of Barteria fistulosa Masters growing in
small forest treefall gaps, but Crematogaster domi-
nate in large clearings (D. McKey, personal obser-
vation). In Barteria nigritana J. D. Hooker, mostly
restricted to light-rich coastal shrub vegetation,
Crematogaster is the only recorded associate. In
Leonardoxa africana Aubrev., Petalomyrmex is
the typical associate of adult trees, and of a large
proportion of juveniles. However, juveniles grow-
ing in deeply shaded sites are usually occupied by
Cataulacus (McKey 1984). The effects of insola-
tion on resource quality can also be apparent within
host species, as in the observation that Polyracliis
species specializing on broad-leaved bamboos build
their pavilions only in sunny areas of bamboo
clumps (Dorow and Maschwitz 1990).

At present, factors underlying interspecific dif-
ferences in the resource demands of ants are poorly
studied. However, just as the evolutionary diversi-
fication of plants has been influenced by tradeoffs
in allocation and life history strategies (e.g.. Grime
1974), similar tradeoffs are likely to have contrib-
uted to a proliferation of divergent ecological tac-
tics in plant-ants (and ants in general, Tschinkel
1991, A. N. Anderson 1991). Included among
these life histories may be: 1) opportunistic (rud-
eral) species with rapid colony growth rates, high
worker turnover, high resource demands, small (or
moderate) colony sizes with correspondingly weak
colony defense, short colony lifespans, and early
reproduction; 2) “tolerant” species with slow-grow-
ing colonies, low worker turnover, low resource
demands, high longevity, deferred reproduction,
and effective defense of the nest site, and 3) com-
petitive species with rapid colony expansion, low
worker turnover, and large, long-lived, aggres-
sively territorial and well-defended colonies. The
evolution of such divergent ecological strategies is
likely to have been influenced also by phylogenetic
constraints, such as preexisting uses of exocrine
glands (Blum and Hermann 1978, Buschinger and
Maschwitz 1984), or the form of the proventricu-
lus, which controls the capacity for and efficiency
of liquid food storage and transport (Eisner 1957).
Such phylogenetic constraints might help to ex-
plain why the competitive rankings and strategies
of ants are often well-defined (though not perfectly
so) at the generic level.

Volume 2, Number 1, 1993

31

Fig. 5. Tiny seedling of Cecropia
“ tessmcnmii ”, whose myrmecophytic traits
appear approximately with the fifth through
seventh leaves past the cotyledon stage, and
when plants are < 10 cm tall. Because of its
extreme morphological similarity to C.
membrcaiacea, C. (prov.) "tessmannii" is still
technically lumped with that species (C. C.
Berg, personal communication). However,
C. membranacea , a pioneer of large, riverine
disturbances, grows more rapidly and ac-
quires its myrmecophytic traits at substan-
tially and significantly later leaf nodes
(Davidson and Fisher 1991).

Other Traits of Weakly Competitive Ants

Appendix 1 reveals numerous exceptions
whereby the generic affiliations of ants are imper-
fect predictors of subordinate or dominant status, as
reflected by pruning behavior and association with
trichome myrmecophytes or uncontested host plants.
Nevertheless, some of these exceptions are consis-
tent with the general principles developed here. For
example, despite their chemical defenses, ants in
some subgenera of Camponotus (especially
Colobopsis and Pseiulocolobopsis) can behave as
subordinates, living secretive lives inside their hol-
low stem nests. Y et Camponotus of this description
occur on a diversity of hosts that lack protective
trichomes and, with one exception, they do not
prune or attack vegetation around their hosts. At
least two factors may explain the capacity of these
species to persist on their hosts. First, major work-
ers use their large and often modified heads to seal

stem entrances effectively and to protect nests from
invaders. Where ants obtain the majority of their
resources from Coccoidea inside stems or domatia
(e.g., under ocrea of Korthalsia), foraging occurs in
seclusion and entails little risk. (A similar explana-
tion may apply to the timid Plieidole colonies from
myrmecophytic pipers and melastomes, which sup-
ply food bodies inside domatia.)

On the other hand, the extrafloral nectar of
Endospennum and the Mullerian bodies of Cecro-
pia , are produced on external plant surfaces. Here,
the exclusivity of ant resources is protected in part
by the temporary nature or temporal pattern of their
production. For example, in the northern coastal
forests of Papua New Guinea, Endospennum labios
Schodde produces almost all of its extrafloral nec-
tar in a brief pulse at about 3:00 AM, likely coincid-
ing with the diel maximum in relative humidity
there (Fig. 6). In contrast to myrmecophytic E.
labios, a myrmecophilic congener, Endospennum
uiedullosum L.S. Smith produces a greater fraction

32

Journal of Hymenoptera Research

of its nectar during other periods of the diel cycle
(D. Davidson, unpublished). Although many Ce-
cropia species release Mullerian bodies slowly all
day long, they also flush large numbers of these
bodies just after nightfall (Davidson and Fisher
1991). Moreover, ants with generalized diets are
usually not attracted to the bodies (Rickson 1977,
D. Davidson, personal observation). Camponotus
associates of Endospenmtm and Cecropia both
forage on leaf surfaces principally at night, and
workers of Anoplolepis (not a plant-ant) can range
freely over Endospennum during daylight hours
(D. Davidson, personal observation). (See also A.
N. Anderson’s [ 1991 ] discussion of noctumality in
Australian Camponotus.) Cladomyrma of
Neonauclea are nocturnal as well (D. Davidson,
personal observation), though the object of worker
foraging on Neonauclea has yet to be identified.
Finally, some plant-ants in the genera My nnelachista
and Allomerus are apparently restricted to their
hosts diumally, but make nocturnal forays to the
forest floor (J. Longino, personal communication).
Together, these observations suggest that competi-
tion may be reduced somewhat at night, though the
nature of any restrictions on nocturnal activity in
dominants is not readily apparent. While activity
schedules of temperate and arid zone ants are
strongly related to diel variation in temperature and
humidity regimes, biotic selection pressures could
be equally important or more important determi-
nants of foraging times in ants of moist tropical
forests.

ANCESTRAL VERSUS MODERN
RELATIONSHIPS

We have argued that the matching of ants and
myrmecophytic plants is convergently alike in dif-
ferent tropical regions, and that this convergence
arises from the presence of similarly preadapted
plants and ants within the respective biotas. In
concentrating on the associations as they exist
today, we have neglected the pathways by which
they may have reached their present form. Ant-
plant symbioses have undoubtedly evolved from
more casual and opportunistic relationships be-
tween plants and ants. In their initial phases, many
of these associations would likely have resembled

modern-day relationships in which plants lack ob-
vious speci al izations for hou si ng ants ( Appendi x I ,
“N” in column 6). Like most other forms of
mutualism (reviewed in Thompson 1982), many
symbiotic ant-plant mutualisms probably began as
parasitisms. What factors may have facilitated the
transition from parasitism to mutualism, and what
character transformations could have accompanied
this change?

For plants hosting ants inside primary domatia
(live stems and intemodes), ancestral relationships
probably consisted of ants tending scale insects
within natural plant cavities or in insect borings
(cf.. Ward 1991). From the start, ants must have
benefitted from access to exclusive resources in
these protected environments. However, to have
remained entirely in the sanctity of the host plant,
ants would have needed a well-balanced diet. Ho-
mopteran exudates contain not only carbohydrates,
but some amino acids and lipids (reviewed in
Buckley 1987), and ant colonies are known to
harvest and eat Homoptera to meet their protein
requirements (e.g.. Way 1954, Pontin 1978). Fur-
thermore, in both New and Old World tropics, as
well as in Australasia, some plant-ants have evolved
means of obtaining added protein and fats from
elaborated calluses or heteroplasias caused by trau-
matic injury to either the inside ( Tetraponera on
African Vitex, Bequaert 1922) or outside of host
plant stems (South American Pseudomyrmex on
Triplaris , and New Guinea Camponotus on
Endospennum [D. Davidson, personal observa-
tion], and possibly Central American Mynnelachista
on Ocotea [J. Longino, personal communication]).
In large part then, coccoid-tending residents of live
stems and cavities could probably have depended
on hosts to satisfy most or all of their nutritional
needs, even from the earliest stages of their rela-
tionships with these plants.

In contrast, the impact of symbiotic ants on their
host plants would have depended on the balance
struck between resource losses to scale insects and
ants, and any anti-herbivore protection the ants may
have originally afforded. Although the majority of
ants would probably have provided at least some
protection against stem and leaf parasites, the
Coccoidea would surely have been a liability.
Substantial carbohydrate losses sustained by the

Volume 2, Number 1, 1993

33

Fig. 6. This large drop of extrafloral nectar was produced in a brief pulse at 3:00 AM on the petiolar nectaries of
Endospenmun labios, at the Christensen Research Station near Madang in Papua New Guinea. (Screenhouse plant
courtesy of M. Jehb.)

plants should have been most debilitating to car-
bon-limited (light-limited) plants. Thus, in habitats
of low light intensity, natural selection on plants
may have acted mainly to exclude both ants and
Homoptera. However, where light was abundant,
the benefits of ant defense could have outweighed
carbohydrate losses (on average). Natural selection
on these plants should have favored attraction of
ants, rather than resistance to them. In this way, the
propensity of ant-parasitized plants to evolve to-
ward myrmecophytism could have been facilitated
by high availability of carbon (light) in relation to
limiting mineral nutrients, and impeded when such
ratios were low. Furthermore, if herbivore pres-
sures are generally more intense in comparatively
productive, sunny environments (see Davidson and
Fisher [1991] for Cecropia), this trend could have
reinforced selection for ant attraction in such habi-
tats.

Although our data set lacks the resolution to test
this hypothesis, the hypothesis is consistent with
the central result of Schupp and Feener’s (1991)
recent survey of the distribution of ant attractants
(EFN’s and pearl bodies) within the flora of Barro
Colorado Island, Panama. While the occurrence of

such rewards was clearly correlated with phylog-
eny, it also appeared to depend on the light environ-
ment. Plant families characteristic of forest light
gaps were overrepresented among ant-defended
families. (See also the frequency of superscripts
“e” and “g” in column 2 of Appendix 1.) Schupp
and Feener hypothesized that the high frequency of
ant defenses among forest gap plants may be ex-
plained by the comparatively low costs of produc-
ing carbohydrate ant rewards in these light-rich
habitats, as well as by the tendency for relatively
continuous growth and leaf production in gap spe-
cies. The latter explanation meshes well with
McKey’s (1989) interpretation of biotic defenses
as an alternative to phenological escape from her-
bivory (i.e., escape from detection, due to variable
and unpredictable new leaf production). Pheno-
logical escape would be unavailable to plants with
continous leaf production.

There are some indications that the absence of
scale insects may be the derived condition in rela-
tionships involving pseudomyrmecines (P. Ward,
personal communication). Thus, although
Coccoidea can be found at the bases of spines on
African and Indian Acacia housing Tetraponera,

34

Journal of Hymenoptera Research

PseudomyrmexAnhsfoited Central American Aca-
cia lack scale insects but supply protein-rich Beltian
bodies. Moreover, the gnawing of internal stem
walls by Tetraponera tessmannii (Stitz)on African
Vitex, to produce tunnels with terminal nutritional
heteroplasias, could have had its origins in the
excavation of pits to increase the feeding efficien-
cies of coccoids, now absent from this system (see
Bailey 1922 for Cuviera).

For plants that continued to be inhabited by ants
and scale insects, natural selection would be ex-
pected to favor a reduction in the ratio of coccoid to
ant biomass. Although many obviously specialized
ant-plants still harbor Coccoidea (Appendix 1),
there is considerable variation across all the ant-
plants in the densities of scale insect populations
(D. Davidson, personal observation). At one ex-
treme are the comparatively unspecialized relation-
ships between Anonychomyrma (previously
Iridomyrmex [Shattuck, 1 992b]) and Crenuitogaster
ants, and a number of pachycaulous understory
New Guinea trees. Here, the biomass and density
of Cryptostigma scales are so great that their popu-
lations may well be limited by either plant re-
sources or the availability of feeding sites (D.
Davidson, personal observation). In contrast, in its
more specialized relationship with Triplaris
americana L., Pseudomyrmex dendroicus Forel
maintains only approximately one scale insect per
leaf junction, and similarly low coccoid densities
are apparent in Cecropia stems inhabited by Azteca
ovaticeps and A. australis.

By what proximate mechanisms might plants
have responded to selection for reducing losses to
Homoptera? For myrmecophilic plants, Becerra
and Venable (1989) have argued that EFN produc-
tion could have arisen as a means of paying ants
directly and eliminating parasitic homopteran in-
termediates. Even if EFN’ s provided ant rewards
comparable to or lower in value than homopteran
secretions, reduced resource handling times might
have induced ants to feed at nectaries and to aban-
don their Homoptera. In turn, plants would have
benefitted from lower rates of infection with ho-
mopteran-mediated diseases and possibly lower
resource losses. One difficulty with applying this
theory to the evolution of myrmecophytes is that it
ignores an important distinction between coccoids

(the usual homopteran associates of plant-ants) and
EFN’s. While EFN’s are relatively promiscouous
resources, accessible to many ants, coccoids tended
inside cavities can be used exclusively by symbi-
otic ant associates. If the latter ants are the most
effective mutualists of the plant, and provide better
protection in the absence of opportunistically for-
aging competitors, selection may favor loss of
EFN’s. There is evidence for such a scenario in
myrmecophytic Asian Macaranga, which, in con-
trast to their non-myrmecophytic congeners, al-
most completely lack EFN’s (Fiala and Maschwitz
1991).

A second difficulty with the hypothesis of
Becerra and Venable (1989) is that it ignores the
possibility that colonies might keep pace with the
added resources (EFN) through short-term rede-
ployment of workers or long-term growth. If so,
ants might continue to tend Homoptera while also
feeding from EFN’s. A plausible alternative hy-
pothesis is linked to the assumption that growth of
ant colonies (like that of plants. Bloom et al. 1 985)
is limited by the ratio of carbon and nitrogen re-
sources. By rewarding ants with abundant carbo-
hydrate but starving them for protein (Carroll and
Janzen 1973), plants might have induced colonies
to consume the majority of their Homoptera. In
support of this argument, Oecophylla longinoda is
known to consume more coccoids when given a
supplemental sugar source (Way 1954). Moreover,
M. Anderson (1991) attributes “switching” be-
tween predation and mutualism in ant-homopteran
relationships (see also Pontin 1958 and 1978) to
changes in the nutritional status of the ant colony.
If homopteran populations are regulated in re-
sponse to ratios of carbon and nitrogen availablity
to ants, colonies might be expected to maintain
their associates at densities which supply these
resources at optimal ratios for colony growth. Cur-
rently, a lack of data prevents further speculation as
to how the relative availability (to ants) of carbohy-
drate and protein might vary with homopteran
densities. Future investigations might profitably
focus on natural or experimentally induced varia-
tion in the relative biomasses of ants and Homoptera
in particular ant-plant systems.

Volume 2, Number 1, 1993

35

PLANT FITNESS IN RELATION TO ANT
SPECIES

In many ecological studies of ant-plant symbio-
ses, investigators have focused principally on the
question of whether or not a given ant associate
benefits its host species. With recently renewed
appreciation for the diversity of ants colonizing
individual myrmecophytes comes the realization
that ants may differ in the protection afforded their
hosts (e.g.,Janzen 1975, Oliveira et al. 1987, Rico-
Gray and Thien 1989, Davidson etal. 1991,Longino
1991a and b, but see Vasconcelos 1990, for a
counterexample), and that associations must be
studied in the context of community-wide interac-
tions. While existing data are too meager to corre-
late protection with specific ant traits, some conjec-
tures are warranted. Rapid colony development,
large colony size, and high levels of worker activity
should enhance host-plant defense. Large insect
herbivores (Coleoptera and Orthoptera) may be
best deterred by active, large bodied workers
(Davidson and Epstein 1989). In contrast, division
of colony biomass among numerous small foragers
may promote fine-grained searching and facilitate
the detection of small prey, for example, lepi-
dopteran eggs (Letourneau 1983, Vasconcelos
1991). Some authors have suggested that small and
timid ants provide little protection against herbi-
vores, but augment the nutrient reserves of their
hosts through deposits of feces and refuse (e.g.,
Janzen 1974b, Beattie 1985). However, at least two
studies have confirmed the effectiveness of small,
docile Pheidole ants in defending against either
insect eggs (Letourneau 1983), or herbivorous lepi-
dopteran larvae (Vasconcelos 1991). While nutri-
ent enhancement has been demonstrated convinc-
ingly in myrmecophytic epiphytes and palms
(Rickson 1979, Rickson and Rickson 1986), tests
have disputed the theory for the symbiotic associ-
ates of Macarcmga (Fiala et al. 1989) and Maieta
(H. Vasconcelos and B. Forsberg, personal com-
munication). On reflection, possibilities for nutri-
ent enhancement are limited by the infrequency of
foraging off the host (Appendix 1, column 4) and,
consequently, by the inability of ants to concentrate
materials from the broader environment.

Two other cases are likely candidates for nutri-

ent augmentation by ants (D. Davidson, personal
observation). First, certain Azteca species center
their carton nests on Tococa and Hirtella species
and contribute to a steady rain of carton and refuse
at the base of the host tree trunk. Second, as a
rheophyte of stream beds and rocky river beaches,
Mynneconauclea strigosa (Korth.) Merrill grows
with its roots anchored in rock crevices. The
Crematogaster ants, which are its dominant associ-
ates in forests west of Lahad Datu, Sabah, pack
refuse and feces into domatia at the distal branch
tips, from which new swollen internodes arise. The
absence of any obvious food reward (including
Homoptera) suggests that ants might leave their
hosts to forage. If such is the case, workers could
concentrate nutrients which enhance fitnesses of
hosts growing in extremely nutrient-poor environ-
ments.

In some cases, myrmecophytism actually con-
tributes to host-plant damage by destructive verte-
brate predators of ant larvae (especially by wood-
peckers [Carroll 1983] and monkeys [Freese 1976,
and J. Terborgh, personal communication, for Ce-
cropia). Damage by primates may be less common
for hosts of ants with powerful stings. First, in
Peruvian Amazonia, Pachycondyla luteola Roger
(the “pungara”) is an obligate symbiont of Cecro-
pia , and its painful barbed stings reinforce verte-
brate learning for a period of seven to ten weeks ( D.
Davidson, personal observation). Avian prefer-
ences for nesting in this (Koepcke 1972) and other
myrmecophytes with stinging ants (Young et al.,
1990) may be at least partly attributable to the
protection which ants afford against primates. Sec-
ond, the Tetraponera of African Barteria fistulosa
also impressed Janzen (1972) as effective deter-
rents of vertebrates, and a black colobus monkey
avoided ant-occupied Barteria , while feeding on an
unoccupied individual nearby (McKey 1974). Gray-
cheeked mangabey monkeys ( Cercocebus albigena
[Gray] ) rip open the branches of this host to prey on
Tetraponera brood, but only if this can be accom-
plished by reaching from a perch in a different tree
(D. McKey, personal observation). Even large,
stinging plant-ants may not affect some vertebrates,
however. Gorillas in the Central African Republic
feed on B. fistulosa leaves and branches apparently
undeterred by healthy, active colonies of

36

Journal of Hymenoptera Research

Tetraponera (M. Fay, personal communication).
Finally, because plant-ants with functional stings
also usually prune the vegetation surrounding their
hosts, crowns inhabited by such ants are usually
sufficiently isolated in forest gaps to avoid the
attacks of primates which visit neighboring trees.

TRENDS IN SPECIALIZATION, SPECIFICITY
AND COEVOLUTION

Because the evolutionary histories of symbiotic
ant-plant systems have been largely independent in
biogeographically disjunct tropical regions (Me Key
and Davidson, in press), intercontinental compari-
sons may provide general insights into the evolu-
tionary dynamics of such systems. In this section,
we discuss evolutionary interactions between ants
and plants, focusing on three questions: (1) Has
specialization of ants and plants followed similar
evolutionary pathways due to parallel and/or con-
vergent evolution in organisms from different con-
tinents? (2) Have evolutionary interactions be-
tween ants and plants contributed to the generation
of diversity in plant-ants and ant-plants? (3) If so,
are these interactions a partial cause of interconti-
nental differences in diversity of ant-plants and
plant-ants? Again we focus mainly on the Ameri-
can and African tropics, whose ant-plant associa-
tions are best known.

The Nature and Causes of Specificity

Symbiotic ant-plant systems are in general more
species-specific than are nonsymbiotic ant-plant
interactions (e.g., Schemske 1983). All tropical
regions contain examples of ant-plants that are
obligately associated with one or a small number of
plant-ant species, which in turn have comparably
restricted host-plant ranges. In such cases, speci-
ficity is doubtless a product of intense evolutionary
interaction. However, much of the seeming speci-
ficity in ant-plant symbioses may be maintained by
ecological processes that require no evolutionary
specialization in ant or plant. We have argued that
characteristic and repeatedly observed plant-ant
matches are the result of species sorting (Jordano
1987) of plants and ants which are mutually pre-
adapted in many attributes related to the interaction
(Davidson et al. 1989 and 1991, Davidson and

Fisher 1991). Driven by the strong competitive
interactions that structure communities of
arboricolous ants, the matching of plants and ants is
determined by plant and ant traits which modify ant
access to plant resources.

In a growing number of ant-plant systems, we
now recognize that seemingly specialized plant-
ants may be capable of living on any of several
hosts, and that many or most myrmecophytes can
persist in association with any of several plant-ants.
Nevertheless, some relationships are more frequent
and/or more durable than others. Understanding
the ecological processes which reduce broad poten-
tial niches of plant-ants and ant-plants to narrower
realized niches is a prerequisite to an evolutionary
investigation of such systems (Davidson and Fisher
1991). First, ecological studies suggest simpler
alternatives which must be excluded before hy-
potheses of evolutionary specialization and coevo-
lution can be entertained. Second, if ecological
causes of specificity can be defined, these will
suggest the likely selective environments in which
any evolutionary specialization may have taken
place. Third, studies of unusual associations may
give clues about the origin of both host-plant speci-
ficity and host switches, which seem to have taken
place frequently in symbiotic ant-plant systems
(Ward 1991).

Evolutionary Specialization of Ants and Plants

If competitive interactions among ants are suffi-
ciently strong and constant, ecological sorting will
produce predictable patterns of ant-plant associa-
tions and a selective environment conducive to
evolutionary specialization (Schemske 1983). Evi-
dence from various tropical regions suggests that
evolutionary specialization of ant-plants and plant-
ants may have been driven largely by competition
among ant species. Even strong pairwise ant-plant
mutualisms, it appears, owe many of their traits to
an evolutionary background of multispecies an-
tagonistic interactions. Whether these character
states were evolved in the context of the symbioses,
or merely fine-tuned from pre-existing traits, often
cannot be argued confidently from existing data.
Nevertheless, many traits of both plant-ants and
their hosts may have been elaborated because of

Volume 2, Number 1, 1993

37

their selective value in the context of symbiotic
association.

Ants . — Because ants actively choose their hosts,
selection should strongly favor specializations for
rapid and efficient host location by queens. In
addition to minimizing exposure to predation and
other environmental hazards, such adaptations could
help to assure priority of access to contested re-
sources. Indeed, competitively inferior ants might
even usurp the hosts of more dominant species by
evolving rapid means of finding and entering these
plants. First, almost nothing is known about the
kinds of information that queens employ to locate
suitable hosts, but a variety of chemical, visual and
other cues may be used at different stages of host
identification. Whatever the mechanisms of host
identification, the abilities of queens to locate and
colonize specific hosts, and their absence from
other hosts and habitats ( Davidson et al. 1 989, Fiala
and Maschwitz 1990, Morawetz et al. 1992), pro-
vide some of the strongest evidence of evolutionary
specialization to the symbiosis. Second, the ex-
treme dorsiventral flattening of the head, thorax
and abdomen of Petalomyrmex queens could have
arisen due to selection for rapid entry of
myrmecophytes in the face of intraspecific and
interspecific competition for hosts (McKey 1991).
Alternatively, however, specialized queen shapes
might have evolved first in generalized stem-nest-
ing ants (Longino 1989b), preadapting such ants to
become specialized plant-ants. Without additional
phylogenetic analysis, adaptations in body shapes
remain indistinguishable from preadaptations.

Once foundresses have safely entered a host,
their success on plants of different growth rate,
maximum size, or lifespan, will likely depend on
key energetic, demographic, and life history fea-
tures of the colony. Intrinsically rapid rates of egg
production and development of incipient colonies
could be favored on fast-growing plants, and
pleometrosis might substitute for this in at least
some ant species (Davidson etal. 1991). However,
before evolutionary specialization can be inferred
from an apparent matching of colony attributes and
plant growth rates, careful phylogenetic analysis
must exclude the alternative hypothesis that ant
traits evolved prior to the origin of the symbiosis. In
the case of the Azteca and Cecropia , this caution is

reinforced by the likelihood that A. ovaticeps and its
relative A. alfari , may have originated from a weedy
species which was typical of second growth vegeta-
tion (Longino 1991b), and whose life histories
could have preadapted it for occupation of rela-
tively fast-growing hosts of riverine succession. In
contrast, A. australis and its relative A. xantliochroa
Roger, are probably derived from carton-building
ancestors (Longino 1991b), whose comparatively
permanent homes may have predisposed them to
evolve life history traits typical of modern-day
descendants on slower-growing and, in most cases,
longer-lived, forest gap Cecropia.

Many or most specialized plant-ants appear to
have been relatively weak competitors, in which
aggressive behavior could well have been maladap-
tive. Nevertheless, within the limited spheres of
their host plants, a number of these ants appear to
have evolved greater similarity to dominants, de-
fending absolute territories defined by the bound-
aries of individual trees. Thus, one scenario appar-
ent in several plant-ant lineages is that of increased
colony size and aggression in response to symbiotic
association with myrmecophytes (e.g., Janzen
1966). For example, the extended colonies of
Myrmelachista ants on pure stands of Tococa
occidentalis (see above) reach an estimated worker
population 1-2 million ants (Morawetz et al. 1992).
Moreover, on Peruvian Cecropia, Pachycondyla
luteola exhibits the largest and most aggressive
colonies achieved by any ponerine ant. Host trees
>30 m tall literally seeth with aggressive, stinging
workers, and populations almost certainly range
into tens or hundreds of thousands of workers (D.
Davidson, personal observation). Even if rigorous
phylogenetic analysis confirms that closest rela-
tives of these ants have much smaller and less
aggressive colonies (as do Pachycondyla sp. nov.
on Panamanian Cecropia hispidissima Cuatracasas,
Davidson and Fisher 1991 ), ecological studies will
be necessary to determine whether the purportedly
evolved demographic responses of P. luteola are
examples of evolutionary accommodation or only
plasticity in colony structure. For colonies nesting
and feeding in the comparative security of
myrmecophytes, increasing worker life spans, and
nest sites which grow, rather than decaying (like
dead twigs), could lead automatically to larger

38

Journal of Hymenoptera Research

worker populations, and greater aggression might
follow as a behavioral response to colony size.
Similarly, the polygyny and/or pleometrosis noted
as typical or occasional in some purportedly highly
specialized plant-ants (Janzen 1966 and 1973,
McKey 1984, Davidson and Epstein 1989,Longino
1989b, Vasconcelos in press) may be a plastic
response to resource availability or competition,
since queen number can vary similarly in other ant
species (Ward 1 989b, Holldobler and Wilson 1990).
Although queens of Azteca australis found their
colonies individually on isolated hosts in small
light gaps, they often cooperate to initiate colonies
on the faster growing plants of riverine distur-
bances, where both rates of food supply and com-
petition from other incipient colonies are greater
(Davidson et al. 1991). At present it is unclear
whether pleometrosis in the latter environment
arises from evolutionary adaptation to competition
or merely from greater numbers of alates produced
and available in that habitat.

Pruning of vines and other vegetation in the
vicinities of hosts is one trait which provides less
ambiguous evidence for evolutionary accommoda-
tion to competition for hosts. Facultative pruning,
requiring the presence of enemy ants, may eventu-
ally prove to be widespread among unspecialized
close relatives of obligate plant-ants. However,
both obligate pruning, and the maintenance of
vegetation-free zones at the host-plant base, appear
to occur predominantly in ants whose highly spe-
cialized diets (Janzen 1966, Davidson et al. 1989,
Fiala and Maschwitz 1990, Morawetz et al. 1992)
and unitary host genera (but see Ward 1991) pro-
vide independent evidence for specialization.

A final category of specialized ant traits may
have little or no relevance to competitive ability but
nonetheless serve as useful indicators of degree of
evolutionary specialization in plant-ants. For ex-
ample, Ward’s (1991) phylogenetic analysis of
pseudomynnecines points to trends for plant-ants
to have reduced eye size and palpal segmentation,
as well as hypertrophied metapleural glands (ex-
cept in cacia-ants). Palpal segmentation is also
reduced in African Engramma (now included in
Technomyrmex , Shattuck, 1992a), in comparison
to other dolichoderines from the Ethiopian region
(Holldobler and Wilson 1990). Reduction in anten-

nal segmentation occurs in some lineages of
Allomerusi Wheeler 1 942), and arboreal stem-nest-
ing Cladomyrma have fewer antennal segments
than do most other formicines (Holldobler and
Wilson 1990). Although 1 0-merous antennae are
characteristic of more generalized Myrmelachista
species (subgenus Hincksidris) which nest in dead
stems, specialized Central American Mynnelachista
plant-ants have antennae with only 9 segments.
Lastly, barbed stings are probably derived in both
Pseudomyrniex ants (Janzen 1966) and
Pachycondyla luteola (D. Davidson, personal ob-
servation). In general, sting defenses may be more
effective against solitary vertebrates than against
social insect enemies (Davidson et al. 1988), and
barbed stings may have evolved under selection to
reinforce learning by vertebrate enemies.

Plants. — In myrmecophytes, domatia and vari-
ous food rewards offer clear support for evolution-
ary specialization, all the more so since the produc-
tion of such structures can entail obvious costs.
Ecological costs of myrmecophytic traits may be
evident in both the presence of ants, as when ant
predators open the nests (see above), and in their
absence, e.g., when herbivores invade and inhabit
foliar or stem domatia (Jolivet 1991, Vasconcelos
1991). Costs are most evident, however, when
myrmecophytic traits are lost in the absence of
symbiotic ants. For example, though Cecropia
peltata L. is myrmecophytic throughout most of its
distribution, conspecifics in Caribbean island popu-
lations lack Mullerian bodies and have trichilia
reduced or absent (Janzen 1973; Rickson 1977).
(Non-myrmecophytic Cecropia schreberiana
Miquel might have been mistaken for C. peltata on
some of these islands [C. Berg, personal communi-
cation].) In more recent history, introduced Cecro-
pia obtusifolia Bertoloni of Hawaii, and C. peltata
imported to both Asia and Africa have either lost
their Mullerian bodies or trichilia, or are polymor-
phic for these characters and exhibit a range of
trichilia sizes (D. Davidson, personal observation,
Putz and Holbrook 1988). Although it could be
argued that such losses are determined environ-
mentally, rather than genetically, African popula-
tions of Cecropia peltata lacked trichilia even when
grown from seed in greenhouses, where progeny of
myrmecophytic congeners from their native habi-

Volume 2, Number 1, 1993

39

tats have never failed to produce trichilia (Davidson,
unpublished).

Selection may also act on plant characteristics
which influence the outcome of ant-ant competi-
tion for the resources offered. In so doing, evolu-
tion might enhance traits which favor the most
effective mutualists (at levels of defense invest-
ment optimal for the plant) over their competitors.
Perhaps most remarkable. Piper ant-plants appar-
ently produce food bodies only when stimulated to
do so by the appropriate Pheidole ants (Risch and
Rickson 1981), or by specialized parasites of the
ant-plant mutualism (Letoumeau 1990 and 1991).
The persistence of Mullerian bodies on Cecropia
trees lacking specialized ants (Rickson 1977, D.
Davidson, personal observation) provides evidence
that these bodies are not recognized by unspecialized
ants as suitable food. Moreover, Mullerian bodies
of at least Cecropia (prov.) “ tessmann', Cecropia
hispidissima , and possibly Cecropia ficifolia
Snethlage appear to have been modified evolution-
arily to favor their usual resident ants (Davidson
and Fisher 1991 ).

In a variety of ways, selection might modify the
quality, rate, timing or position of the food reward
to encourage either fine-grained or coarse-grained
foragers, large or small workers, and aggressive,
energy-intensive competitive dominants or timid,
energy -conservative subordinates (see above). As
an extreme example, plants which provision ants
with complete diets may facilitate the persistence
of weakly competitive species, whose foraging can
then be restricted to the host itself (Appendix I,
column 4). At least some species of Triplaris
induce fine-grained foraging by ants with highly
specialized foraging behaviors. These hosts pro-
duce pearl bodies which are unique in their yellow
color (perhaps indicative of some distinctive nutri-
tional quality) and. are distributed in patches on
adaxial leaf surfaces. Perhaps pre-adapted for this
behavior by prior dietary specialization on pollen
and fungal spores (Wheeler and Bailey 1920), the
Pseudomyrmex residents of these myrmecophytes
accumulate these tiny food bodies on their append-
ages while constantly traversing leaf surfaces. They
groom the material frequently onto their sting
sheaths, which serve as storage sites until workers
return to their nests (Davidson et al. 1988).

In a number of ant-plant genera, food rewards
for ants are often produced in more localized and
defensible sites on true myrmecophytes than on
myrmecophilic relatives with more promiscuous
rewards. Thus, in the Endospennwn of New Guinea
(Airy-Shaw 1980), myrmecophilic E. wedidlosum
has moderately sized EFN’s scattered across abaxial
leaves along primary and secondary veins. Petiolar
nectaries are only slightly larger. In comparison,
myrmecophytic congeners have greatly enlarged
petiolar EFN’s and all other EFN’s greatly reduced
in size and number. Similarly, in the genus
Macaranga , at least some myrmecophilic species
have scattered pearl bodies used by a number of
unspecialized ants (D. Davidson, personal observa-
tion in New Guinea), whereas the most highly
evolved myrmecophytes restrict access to food
bodies by hiding them beneath recurved stipules.
In incipient myrmecophytes, A/. hosei King ex Hk.
f. and M. pruinosa (Miq.) Muell.Arg., whose stems
are not naturally hollow and are only partially
occupied, accessibility of food bodies appears to be
intermediate (Fiala et al. 1991). Thus, although
food bodies are locally concentrated on stipules, the
stipules are horizontal, leaving them exposed.
Experimental studies might focus profitably on the
outcome of ant-ant competition in relation to the
spatial patterning, accessibility and defensibility of
ant rewards. Similar relationships are well ac-
cepted for other plant-animal mutualisms (e.g.,
Feinsinger and Colwell 1978).

Restrictive entrances to domatia (Fig. 7) may
render these structures more readily habitable by
some ants than others, as well as limiting access to
stem-dwelling Coccoidea. Prostomas of myrme-
cophytic Leonardoxa are matched to the shapes and
sizes of their associated ants (McKey 1991),
Urticating hairs on the prostoma of Cecropia (prov.)
“ tesswannii" favor large-bodied queens of
Pachycondyla luteola over smaller-bodied Azteca
queens (Davidson and Fisher 1991). In general,
neither the selective effects of these traits on differ-
ent ant associates, nor their consequences for plant
fitness are well documented. Nor is it often clear
where “preadaptation” stops and adaptation be-
gins. For example, the thin pith cavities of
myrmecophytic Vitex lianes are easily exploited by
the plant’s specialist associate, the slender

40

Journal of Hymenoptera Research

Tetraponera tessmannii , but not by stouter ants of
similar body length. At present, however, there is
no evidence to suppose that either plant or ant has
evolved to produce or enhance such a match. Even
the long, elliptical prostoma of Leonardoxa africana ,
matched to the flattened queens of Petalomynnex
(McKey 1991), might be explained as preadapta-
tion. As in numerous other ant-plants with stem-
domatia, this myrmecophyte’s prostoma occurs at
the node, opposite the leaf insertion, where a reduc-
tion in xylem leaves the stem wall relatively thin
(Bailey 1922). In future studies, both field experi-
ments and careful phylogenetic analyses of plant
and ant lineages will be required to determine how
frequently myrmecophytes may have evolved to
influence ant-ant competition.

Limits to Specialization

The forces leading to specialization in ant-plant
symbioses are both clear and consistent with theo-
retical arguments predicting greater specialization
in mutualistic systems where strong antagonistic
interactions occur among competing mutualists
(Law and Koptur 1986). What factors then limit
species specificity and account for the persistence
of systems in which multiple ants coexist on the
same host, or a single ant occupies several hosts?
What are the limits to specialization? First, the
matches produced by ecological sorting do not
necessarily result in mutualistic interactions. A
plant may be fortuitously 'preadapted’ to harbor a
persistent parasite, as well as an effective mutualist.
Depending on the match, an association might
engender strong reciprocal specialization (when
most effective mutualists are paired), asymmetrical
specialization, or even antagonistic interactions in
which specialization in ants and plants proceeds in
opposite directions.

Even when ant-plant associations are funda-
mentally mutualistic, there may be both genetic and
ecological limits to specialization and coevolution
(Schemske 1983, Kiester et al. 1984, Howe and
Westley 1988). The nature of any genetic con-
straints is purely a matter for speculation. By and
large, we do not know the extent of heritable
variation for relevant ant and plant traits, nor whether
such variation might limit specialization. Like-

wise. population structure of ant-plants, and espe-
cially that of plant-ants, is too poorly understood to
support much discussion of how specialization and
species origination might take place in these sys-
tems. Since sexual selection can drive rapid evolu-
tionary specialization and coevolution in mutual-
ists, added information on ant mating sites and
behaviors or data from genetic markers might be
especially interesting in helping to determine
whether mating could be non-random with respect
to the host species where alates originated.

More can be said about potential ecological
limits on the intensity of selection for specializa-
tion. Most significantly, the outcome of an ant-
plant interaction may often depend not only on the
specific identities of associates but also on habitat
type and plant size. As summarized above, habitat
may influence the match between plants and ants
through both ecological and evolutionary variation
in rates of resource supply to ants. The effects of
habitat heterogeneity could also be mediated through
other mechanisms that are still poorly understood.
For example, on isolated plants, or where nutrient
poverty limits productivity and alate production,
low frequencies of host plant colonization may
reduce the intensity of ant-ant competition for hosts
(Vasconcelos, in press, D. Davidson, personal ob-
servation). Herbivore pressures on at least some
myrmecophytes appear to differ with habitat and
plant size (Davidson and Fisher 1 99 1 , Janzen 1 974a,
Letoumeau 1983), as does the probability that
overgrowing vines will threaten both the host and
resident ant colony (Rickson 1977, Davidson and
Fisher 1991). Perhaps also varying with habitat are
the densities of queen and brood parasitoids, which
either kill incipient colonies, or prolong their devel-
opment (Davidson and Fisher 1991). Finally, the
outcome of competition among ants for host plants
may be influenced by habitat-correlated physi-
ological effects on colony development. In Azteca
ovaticeps , queen mortality prior to first worker
production is much higher on shaded hosts at the
forest edge than on hosts of large, sunny and hot
riverine disturbances (Davidson et al. 1991). In
both the Azteca of South American Cecropia and
myrmelachistines of African Leonardoxa , inter-
specific variation in queen color correlates with
habitat in a manner consistent with the hypothesis

Volume 2, Number 1, 1993

41

Fig. 7. Restrictive entrance to domatia of the African myrmecophyte Leonardoxa africana (Baill.) Aubrev.
(Fabaceae: Caesalpinioideae). The plant's mutualistic ant associate, Petalomyrmex phylax Snelling, makes these
slit-like entrances at the site of the prostoma, which is of similar shape. The entrance allows access by the specialized
dorsoventrally flattened foundresses of P. phylax, but not by other ants of similar size. Workers of P. phylax are
of normal shape, but can easily pass through these entrances because they are much smaller than dealate queens of
Petalomyrmex or workers of other ant species associated with the plant.

that black queen coloration could be adaptive on
fast-growing hosts, possibly because of a positive
effect on physiological rates. Queens are black in
A. alfari , which dominates Cecropia of roadsides
and pastures in many disturbed regions, and yel-
lowish brown in A. ovaticeps , the typical resident of
fast-growing riverine Cecropia (Longino 1989b).
Occurring mainly on Cecropia of small forest light
gaps, A. australis has yellow queens, and may have
comparatively slow rates of egg-laying (Davidson
et al. 1991). Similarly in Leonardoxa , black-bod-
ied queens (and workers) of Aphomomynnex tend
to occur in more exposed riverine situations, whereas
reddish yellow Petalomyrmex are typical of more
shaded forest understory.

Within myrmecophyte species, host size-de-
pendent variation in the relative abundances of
alternative plant-ants may be determined in some
cases by the match between colony resource de-
mands and rates of resource provisioning by the
plants. However, other causal mechanisms might
also produce correlations between plant sizes and
the identities of ant inhabitants. For example, such
correlations could occur if ant species differed in
the capacity to protect their hosts from herbivory
(suggested by Longino 1991a and b, for Central
American Azteca on Cecropia). Additionally, a
form of ecological succession may take place, with
regular changes in ant inhabitants through indi-
vidual plant lifespans. Turnover of ant species

42

Journal of Hymenoptera Research

through time has been observed on hosts in the
genera Acacia (Janzen 1 975), Leonardoxa (McKey
1984), Taclngcili (Benson 1985), and Maieta
(Vasconcelos 1990). Just as successional mecha-
nisms may vary across plant communities (Connell
and Slatyer 1977), they may also vary across ant-
plant systems. One possible explanation for spe-
cies replacements is that early colonists are eventu-
ally replaced by superior competitors (Janzen 1 975,
McKey 1984, Davidson et al. 1989). In this
context, coexistence of multiple ant species on a
single host population requires that competitive
abilities be inversely proportional to colonizing
abilities, with poor competitors making a living as
“fugitive species”.

Even in the absence of direct interspecific inter-
actions, disparate ant life histories might lead to
successional changes among the ants of individual
hosts. On Central American Acacia, for example,
Pseudomyrmex nigropilosa Emery is an opportu-
nistic colonist and short-term resident after prior
residents have died from fire and other causes
(Janzen 1975). A similar mechanism has been
proposed by Vasconcelos ( 1 990) to account for the
coexistence of Pheidole winutula Mayr and
Crematogaster sp. on Maieta guianensis Aublet
near Manaus, Brazil . Although the two ant species
provide equivalent protection for their hosts, the
frequency of Pheidole occupancy increases with
plant size. Comparatively early death or desertion
of hosts by Crematogaster (for unknown reasons)
leaves plants to be colonized again. Whatever the
average relationship between the colonizing abili-
ties of the two ants, larger plants should eventually
accumulate Pheidole colonies, due to the frequent
abandonment of hosts by Crematogaster.

To summarize, both the species composition of
ant-plant symbioses, and the fitness consequences
of particular associations, can vary markedly in
space and time. Just as such inconsistencies are
postulated to have limited evolutionary specializa-
tion in non-symbiotic ant-plant relationships
(Schemske 1983, Beattie 1985), they have likely
been the predominant obstacles to the evolution of
species-specificity in symbiotic associations.

Evolutionary Dynamics of Ant-plant Symbiosis

Given these limitations to species specificity,
what are the implications for coevolution? Coevo-
lution has two aspects. The first is co-accommoda-
tion, reciprocal evolutionary responses of interact-
ing organisms (Mitter and Brooks 1983). Co-
accommodation is most easily recognized when it
involves functionally matched characters of associ-
ated organisms or coupled character coevolution
(Schemske 1983). Several ant-plant systems in
both Africa and South America offer examples
suggestive of reciprocal specialization of function-
ally matched characters in plants and associated
ants. In this category are matches between the
dimensions of ants and the prostomas of their plant
associates (McKey 1991, Davidson and Fisher
1991), and between food provisioning by plants
and the foraging and pruning behaviors of their ants
(Davidson et al. 1988). Though suggestive, the
data are not usually sufficient to pass arigorous test,
especially in view of our poor knowledge of phylo-
genetic relationships (McKey 1991).

The second aspect of coevolution is association
by descent (Mitter and Brooks 1983). If ant-plant
relationships have persisted and diversified as the
associated lineages underwent successive specia-
tional events, their phylogenies should be congru-
ent. If, on the other hand, events such as host-
switching and secondary exploitation of preexist-
ing ant-plant mutualisms are frequent, there will be
no close correspondence between ant and plant
phylogenies. Interspecific hybridization of plants
and/or ants will produce yet a third pattern, reticu-
late evolution. Janzen (1974a) concludes (without
rigorous phylogenetic analysis) that the neotropical
ant-acacias do not form a tight phyletic group, and
postulates that one species may capture ant-adapted
traits from another via introgression. Ross (1981)
came to similar conclusions regarding African ant-
acacias. Aside from the two groups of Acacia, there
is little information to evaluate the possible role of
hybridization in the diversification of ant-plants.
Moreover, Janzen’s observations might beexplained
alternatively by genotype-environment interactions.
Thus, evolved associations of ants with one acacia
lineage could have increased the selection intensity
for myrmecophytism in other (possibly preadapted)

Volume 2, Number 1, 1993

43

lineages, perhaps because ants occasionally colo-
nized these unspecialized hosts.

Of the relatively small number of taxa which
have produced modest to extensive radiations of
ant-plants or plant-ants, taxonomic uncertainty pre-
cludes any examination of the question of associa-
tion by descent in all but a few cases. And in no case
do we have equally robust phylogenies in both ants
and plants. By far the best example is Ward’s
(1991) study of associations between plants and
pseudomyrmecine ants, represented by Pseudo-
myrmex and Myrcidris (Ward 1990) in the
Neotropics, and by Tetraponera in Africa, Asia and
Australasia. Specialist plant-ants appear to have
arisen at least 1 2 times in this sub-family, on a wide
range of hosts. Most of these events have produced
only one or a few species of plant-ants, associated
with a comparably small number of host species.
Such small radiations offer limited opportunity for
association by descent. In some cases, apparently
secondary pseudomyrmecine colonizations of pre-
existing ant-plant mutualisms have given rise to a
small number of species on Cordia, Pleurothyrium
and possibly Cecropia (Ward 1991), all of which
are predominantly associated with other ants
( Allomerus , Mynnelachista and Azteca, respec-
tively).

The hosts of pseudomyrmecines do include,
however, three plant genera with large numbers of
ant-plant species. Each of these (neotropical Aca-
cia, Tachigali, and Triplaris) is associated with a
different monophyletic group of Pseudomyrmex.
Do these more extensive radiations offer evidence
of association by descent? Ward (1991) concludes
that at the species level, they do not. First, within
each of these groups there is no pairwise specificity
of ant and plant species. Not surprisingly, there is
no clear pattern of cospeciation. Although in each
of these three cases, the plant lineage seems to have
evolved in concert with the ant lineage, the pattern
of associations suggests host shifts within a taxo-
nomically restricted guild of ants and plants, rather
than cospeciation. Furthermore, each of these plant
groups also harbors ants from at least one other
lineage of Pseudomyrmex. Even in these extensive
radiations from associated ancestors, coevolution
seems to have been diffuse, corresponding to the
guild coevolution or ecological replacement hy-

potheses (Howe and Westley 1988), rather than to
a hypothesis of pairwise coevolution.

Relationships of various plant-ants to neotropical
Cecropia paint a somewhat similar picture. Within
ponerines of the genus Pachycondyla , four prob-
able Cecropia specialists represent at least three
separate origins of specialization on this host ge-
nus. Independent origins include species near both
P. villosa (Fabricius) and P. unidentata Mayr (J.
Longino, personal communication) as well as
Pachycondyla sp. nov. in Panama. Of these, the
first two species appear to be stem parasites. Their
small, secretive colonies show little activity on host
surfaces, though workers of at least the species near
P. villosa harvest Mullerian bodies and locate en-
trances at prostomas ( J. Longino, personal commu-
nication). At present, no data suggest specificity of
host range within the genus Cecropia. In contrast,
Pachycondyla sp. nov. appears to have a highly
specialized relationship with C. hispidissima , which
produces especially large, hard and purple Mullerian
bodies (Davidson and Fisher 1991 , B. Fisher, per-
sonal communication). A close phylogenetic rela-
tionship between this ant and the Peruvian P. luteola
cannot yet be ruled out (W. L. Brown, personal
communication). Colonies of the latter ant occur
only on C. (prov.) “ tessmannii' , whose relation-
ship to C. hispidissima is currently uncharacterized.
The affiliations of Pachycondyla sp. nov. and P.
luteola with their respective hosts are the most
likely candidates for pairwise coevolution between
ants and Cecropia trees, and the evidence is still
weak. Even if ant and plant phylogenies turn out to
be congruent here, and if speciation events are
determined to have been synchronous in ant and
plant lineages, any postulated cospeciation would
appear to have been minimal, based on the small
number of Pachycondyla specialized to Cecropia.

Three other ant genera provide support for mul-
tiple independent colonizations of Cecropia. The
genus Camponotus includes at least two host gen-
eralists, C. balzani Emery in southeastern Peru, and
an unnamed species of Camponotus sub-genus
Pseudocolobopsis in northern Peru( Davidson, un-
published; R. Snelling, personal communication).
Multiple radiations of specialized A zteca (Longino
1989b, 1991a and b) were mentioned above. Al-
though phylogenies are not yet defined within ei-

44

Journal of Hymenoptera Research

ther ant genus, the overlapping and generalized
host ranges of closely related ant species argue
against cospeciation as the major mechanism by
which diversity is generated. Finally, at least one
Crematogaster species (near C. cun’ispinosa Mayr,
J. Longino, personal communication) appears to be
a specialist on Cecropia in northeastern Peru (vie.
Genaro Herrera), but inhabits at least several differ-
ent hosts within the genus (D. Davidson, personal
observation). With specialized symbionts repre-
senting four of the five sub-families of plant-ants,
and multiple origins within at least three ant genera,
Cecropia presents a strong case for the ease with
which taxa of generalized stem-nesting ants have
colonized myrmecophytes over evolutionary time.

Like Pseudomynnex and Tetraponera, many
other plant-ant genera are associated with numer-
ous, unrelated plant hosts (Appendix 1). Of 31
plant-ant genera (including various subgenera of
Camponotus ), only 1 1 are known from a single host
genus, and three of these are records for species
whose specialization as plant-ants (column 7) re-
mains in doubt. As in pseudomyrmecines, these
broad generic host ranges are probably due both to
multiple independent origins of the plant-ant habit
within the ant genus, and to secondary colonization
of additional hosts by plant-ant species. However,
the taxonomic information necessary to distinguish
between these possibilities is lacking. Allomems is
a particularly intriguing case. All known species
are specialist plant-ants. Unless we assume that
non-specialist Allomerus once existed but are now
all extinct (the genus has no fossil record [Holldobler
and Wilson 1 990] ), then the host range of this genus
(seven plant genera in five families) is due to
secondary colonizations and host shifts.

Perhaps the clearest evidence against
cospeciation is offered by those cases in which a
prerequisite for cospeciation, host-specificity, is
not fulfilled. Several plant-ant species are associ-
ated with two or more quite unrelated hosts. At
least three specialist plant-ant species of
Pseudomynnex occupy more than one plant genus
(Ward 1991), with P. vidiius F. Smith recorded
from 5 genera in as many families. Aphomomynnex
afer Emery is associated with Vitex (Verbenaceae)
and Leonardoxa (Fabaceae) (R. Snclling, personal
communication). Technomyrmex (formerly

Engramma) kohlii is associated with five genera
(Cola, Scaphopetalum, Cantliimn, Diospyros and
Delpydora ) belonging to four families (Bequaert
1 922; R. Snelling, personal communication). These
appear to be cases in which secondary colonization
of ant-plants has occurred several times.

At least one other case, however, does suggest
association by descent. African Leonardoxa in-
cludes two myrmecophytes, which cladistic analy-
sis has shown to be sister species (McKey 1991).
They are inhabited by Aphomomynnex afer and
Petalomynnex phylax Snelling, respectively, the
only two African representatives of the formicine
tribe Myrmelachistini. Though these two ants are
obviously closely related (Agosti 1991), further
taxonomic work will be required to determine
whether they are sister species or relicts of formerly
diverse genera in which all congeners have gone
extinct.

Habitat Specialization and the Generation of
Diversity in Ant-plant Synibioses

Our analysis indicates that cospeciation in lin-
eages of plants and of host-specific ants has been
infrequent at best. Ant-plant pairs may be co-
evolved, but associations seem to be shuffled or
broken frequently, rather than diversified in con-
cert via cospeciation. Pairwise coevolution thus
can account for little of the diversification of these
symbioses. How then have symbiotic ant-plant
associations diversified? Mounting evidence sug-
gests that evolutionary interactions in these sys-
tems, in both Africa and the Neotropics, correspond
more closely to two other models of coevolution,
not mutually exclusive, the guild coevolution hy-
pothesis and the ecological replacement hypothesis
(Howe and Westley 1988). These hypotheses en-
visage diffuse evolutionary interactions among
sympatric guilds of associated organisms. Specia-
tion may be accompanied by shifts in patterns of
host associations, producing new mixes and
matches. In these guilds, one member may replace
another as the predominant associate of a particular
member of the other guild. Guilds are also open.
New ants may colonize pre-existing ant-plant
mutualisms, perhaps displacing or completely re-

Volume 2 , Number 1 , 1993

45

placing other ants, and new plants may join a guild
of ant-plants.

We postulate that habitat-dependence in the
outcome of different ant-plant interactions has been
the principal force driving host shifts and ecologi-
cal replacements within these guilds. Thus, the
main obstacle to species-specificity and pairwise
coevolution of ants and plants has at the same time
facilitated diversification by other mechanisms.

Host plant quality, as recognized by ants, may
vary more with habitat than with host species. Thus,
Janzen (1966) has called attention to disparities in
the habitat associations of P. nigrocinctus (Emery)
and P. spinicola Emery (= P. ferruginea), though
the two closely related (Ward 1991) species coexist
locally. In some parts of their ranges, these two
species also coexist with P.flavicornis F. Smith (=
P. belti), which has yet a different pattern of habitat
association (Janzen 1983). In another example,
distributions of obligate Cecropia ants, both within
and across genera, are usually more responsive to
habitats than to host species (Harada and Benson
1 988, Longino 1 989b, 1991a and b, Davidson et al.
1991). As is the case for acacia-ants, the conse-
quent mixing and matching of ants and Cecropia
species may favor diffuse rather than pairwise
coevolution. Likewise, effects of different ants on
plant fitness may vary with habitat, for example, if
the quality of defense against herbivores mattered
less under favorable than unfavorable resource
regimes.

Thus, genetic differentiation may be associated
more frequently with habitat specialization, both in
plants (Davidson and Fisher 1991 ) and in ants, than
with specific identities of associates. However,
habitat-dependence may still drive a type of
cospeciation. For example, a plant and an associ-
ated ant may have parallel genetic responses to
environmental variation, both of them diverging
from conspecifics in a different habitat. Or, genetic
differentiation in one symbiont, driven by habitat
specialization, may induce divergence in its associ-
ate (Thompson 1987). The likelihood of such
events, in which both ant and plant remain associ-
ated while undergoing habitat-related divergence,
may depend on guild diversity. Thus, when an ant-
plant colonizes a novel environment, poor success
of the usual ant associate certainly provides selec-

tive pressure for adaptation of the ant to the new
habitat. But it also provides opportunities for the
establishment of other ant species. The richer the
local guild of plant-ants, the greater the likelihood
that a member of the guild will establish success-
fully, replacing the usual associate and preventing
its specialization for the novel habitat. In depauper-
ate guilds, preadapted ants are fewer, and the usual
associate may be more likely to persist and adapt to
the novel environment. A possible example is the
relationship between Leonardoxa spp. and their
Petalomynnex and Aphomomyrmex ants. Plausi-
bly a case of cospeciation, this system involves a
small number of ant and plant species (McKey
1991 ). Neither the two plants nor the two ants ever
occur sympatrically , and few other myrmecophytes
and domatia-inhabiting ants share their habitats.
Perhaps pairwise specificity and cospeciation are
more likely to occur in modest and geographically
limited radiations such as these, where taxonomic
poverty of sympatric guilds of ant-plants and plant-
ants offers little scope for host-switching and sec-
ondary colonization. The latter processes may
dominate in species-rich guilds. If our hypothesis
is correct, it would suggest that diversity begets
diversity due to genotype-environment interactions
in tropical ant-plant symbioses.

EVOLUTIONARY TRENDS IN SPECIES
REPLACEMENTS WITHIN PLANT-ANT GUILDS

Host-switching, secondary colonization, and eco-
logical replacement seem to be the predominant
modes by which ant-plant associations are modi-
fied. Once a new association is forged, it is likely
to engender selection on one or both partners, and
to give rise to evolutionary diversification. But
how do new associations form and spread? What is
their effect on preexisting associations? Can we
recognize patterns in the radiation of plant-ants and
ant-plants? Once again, it may be possible to
understand the evolutionary dynamics of ant-plant
associations in the context of competitive interac-
tions among ants, and habitat-dependence in the
outcome of ant-ant and ant-plant interactions. While
many species replacements may have occurred
without perceptible trace, contemporary systems in
which ant-plants are associated with multiple unre-

46

Journal of Hymenoptera Research

Table 2. Earliest fossil records of ants for genera (worldwide) in which specialized plant-ants
have evolved (summary excerpted from Holldobler and Wilson 1990): A = Arkansas amber
(USA, middle Eocene); Ba = Baltic amber (northern Europe, early Oligocene); Br = Britain
(Oligocene); Do = Dominican amber (Dominican Republic, late Miocene 3 ); F = Florissant
shales, Colorado, USA, Oligocene); Sh = Shanwang shales (China, Miocene); Si = Sicilian
amber (Sicily, Miocene).

Sub-family and tribe

Genus

Earliest fossil find

PONER1NAE

Tribe Ponerini

Pacliycondyla

Early 01igocene Ba ' r-

PSEUDOMYRMEC1NAE

Mwcidris

No fossil record

Pseudomyrmex

01igocene DF

Tetraponera

Early Oligocene 63

MYRMICINAE

Tribe Cephalotini

Zacryptocems

Late Miocene 0

Tribe Crematogastrini

Crenuitogaster

Miocene Sl

Tribe Leptothoracini

Leptothorax

Early 01igocene Ba

Tribe Pheidolini

Pheidole

01igocene D ' F

Tribe Solenopsidini

Allomenis

No fossil record

Solenopsis

Late Miocene 0

Tribe Tetramoriini

Tetramorinm

No fossil record

Tribe Dacetini

Stmmigenys

No fossil record

Tribe unclassified

Cataidacus

Miocene Sl

Podomyrma

No fossil record

Atopomyrmex

No fossil record

DOL1CHODER1NAE

Tribe Tapinomini

Anonychomyrma

No fossil record

Axinidris

No fossil record

Azteca

Early Miocene 0

Tapinoma

Miocene Sl

Technomyrmex

Miocene Sl

FORMIC1NAE

Early Ohgocene Ba ’ Si

Tribe Plagiolepidini

Plagiolepis

Tribe Myrmelachistini

Aphomomyrmex

No fossil record

Ctadomynna

No fossil record

Myrmelac/iista

No fossil record

Petalomyrmex

No fossil record

Tribe Camponotini

Camponotus

Early Oligocene 8 J,Sl

a

Note added in proof. Although Holldobbler and Wilson (1991) date the Baltic amber as late Oligocene,
more recent work summarized by Kirsshna and Grimaldi ( 1991 ) suggests an earlier estimate. We use
the latter date because it is conservative in relation to our hypothesis.

Volume 2, Number 1, 1993

47

lated plant-ants may offer examples of species
replacements in progress. The various ant associ-
ates of myrmecophytes usually occupy different
places in a competitive hierarchy. An understand-
ing of their competitive relationships, and how they
coexist today, should provide insights into the
ecological mechanisms that have driven their evo-
lutionary histories.

Without more phylogenetic evidence than exists
today, we have only a snapshot of a process in
motion, and cannot know its direction with cer-
tainty. Nevertheless, we attempt a provisional
distinction between original associates and second-
ary colonists of several ant-plant associations. First,
we focus on two ant lineages which seem to have
played predictable and frequent roles in the eco-
logical replacement of primary associates. We then
examine likely causes of such pattern, based on
what is known of the biology and competitive
relationships of the ants involved. Generalizing
from these examples, and referencing the fossil
record, we propose a hypothesis of taxonomic
progressions within lineages of plant-ants. This
hypothesis, combined with information on the geo-
logical history of mesic-forest environments in
different tropical regions, leads to new interpreta-
tions of intercontinental differences among ant-
plant symbioses.

Directionality of Species Replacements

The primary and secondary associates of many
myrmecophytes can be very difficult to distinguish
(Ward 1991 ). Nevertheless, patterns in the biogeo-
graphic and taxonomic distribution of host associa-
tions in some ant-plant systems suggest that
myrmecophytes have been colonized recently by
unspecialized arboreal ants or by host-shifting plant-
ants, resulting in partial or complete replacement of
a prior ant associate. In none of the examples that
follow is the evidence for directionality conclusive.
Nevertheless, taken together the evidence is strongly
suggestive, and the approach has enabled us to
propose testable hypotheses and to define critical
points where data required to test these hypotheses
are lacking.

Crematogaster as Secondary Associates of
Myrmecophytes. — Several ant-plant relationships

provide indications that ants of the genus
Crematogaster have partially or completely re-
placed prior ant associates of the host plant. First,
the pattern of ant associations with the two African
Barteria species suggests that ancestral host rela-
tionships may have involved Tetraponera ants. For
T. aethiops (F. Smith) and T. latifrons (Emery), two
host-specific associates of B.fistulosa Mast. (Janzen
1972), taxonomic isolation from other sections of
the genus suggests comparatively ancient origins
for the association (P. Ward, personal communica-
tion). Tetraponera has not been found to inhabit the
other described species of Barteria , B. nigritana
Hook, f., which instead houses an apparently
unspecialized Crematogaster. The latter associa-
tion may have arisen via secondary colonization of
hosts in the more disturbed, light-rich, coastal scrub
sites frequented by this plant species. Interestingly,
while B. fistulosa is occupied by its specialist
Tetraponera in forest light gaps, it too occurs with
unspecialized Crematogaster in large, human-made
clearings in coastal forests of Cameroon (D. McKey,
personal observation).

Second, although Crematogaster spp. are pres-
ently the numerically dominant associates of East
African ant-acacias, Tetraponera ants may have
been the original inhabitants. Invasion of East
African acacias by Crematogaster , which gener-
ated two new specialists on Acacia , may have
largely pushed the weakly competitive
pseudomyrmecine into marginal high-elevation sites
(Hocking 1970). At lower elevations (ca. 900 m), T.
penzigi (Mayr) appears to be competitively subor-
dinate to Crematogaster mimosae (Santschi) and
C. uigriceps Emery, and has exclusive possession
of only 0.7 % of the trees. At higher elevations, it
maintains control of up to 8.5 % of host trees. In
sites where it cooccurs with the two Crematogaster ,
the pseudomyrmecine appears to persist mainly in
unoccupied parts of Crematogaster-occupied trees.
There it ensures exclusive occupancy of stipular
swellings by boring entrance holes too small to
accomodate Crematogaster, and by plugging or
protecting these entrances with carton baffles.

Asian Macaranga may be another case where
contemporary numerically dominant Crew wrogfl.ster
ants have largely replaced the original inhabitants.
Poorly known associations occur between two

48

Journal of Hymenoptera Research

Camponotus species [provisionally subgenus
Colobopsis ] and both Macciranga griffithiana M. A.
and Macarcinga puncticulata Gage (Fiala et al.
1990). Each of these hosts grows principally in
swamplands (Whitmore 1973 and 1975), marginal
habitats where rates of plant growth and supply of
ant resources are likely to be reduced. Finally, one
Crematogaster lineage may also have replaced
another. Thus, Macarcmgci hosts in some undis-
turbed primary forests are occupied by a species
with black workers and 11 -segmented antennae,
whereas hosts of forest and riverine edge typically
contain any of an unrelated complex of species with
yellowish workers and 10-merous antennae (D.
Davidson, personal observation). Despite habitat
segregation under natural conditions, a mixture of
the two ant lineages occurs in the extensive
Macarcmgci forests left after logging. Clearly, in
view of the habitat specificity of both
myrmecophytes and their ants, the rapid conver-
sion of primary forests can be expected to alter
these symbiotic associations greatly in future years.

In the Neotropics, unspecialized Crematogaster
are recorded as clear newcomers and secondary
associates of several older ant-plant relationships,
including those between Pseudomyrmex and
Triplaris (Davidson et al. 1988; Oliveira 1987),
Pseudomyrmex and Acacia (Janzen 1983), and
Azteca and Zacryptocerus with Cordia alliodora
(R. Carroll, personal communication). These Neo-
tropical examples include no obvious case in which
colonization by Crematogaster has led to complete
replacement of a prior associate, and American
Crematogaster have only rarely evolved into spe-
cialist plant-ants. Included in the latter category are
only the Crematogaster cf. victima of many
neotropical leaf-pouch myrmecophytes, and a de-
rivative of the opportunistic and widespread C.
curvispinosa on Cecropia in northeastern Peru (D.
Davidson, personal observation).

Azteca as Secondary’ Associates of Neotropical
Myrmecophytes. — In species richness, Azteca are
the preeminent competitive dominants among New
World plant-ants (Appendix 1 ), and play ecological
roles analogous to those of Crematogaster in many
Old-World systems (Carroll 1983). Like
Crematogaster, they may be displacing subordi-
nate species in many relationships. For example.

both Crematogaster and Azteca ants displaced
Pseudomyrmex deudroicus when permanent wire
bridges were made between the host trees and
neighboring vegetation (Davidson et al. 1988).
Moreover, as we also suspect for Old-World
Crematogaster , some displacements of primary
associates by Azteca may have been so thorough
that distinguishing contemporary from prior asso-
ciations is fraught with uncertainty. For example,
ants of the genus Azteca are the numerically pre-
dominant associates of myrmecophytic Cecropia
today, but associations of Cecropia with other ants,
such as Camponotus and Pachycondyla, may be
older. Each of these latter genera includes species
which are Cecropia specialists, and in both cases
ongoing competition with Azteca may exclude them
from riverine and other riparian habitats, where
Cecropia is most abundant and fast-growing (see
above, Davidson and Fisher 1991).

Replacements may also be occurring within the
genus Azteca. In Amazonian Peru, Azteca ovaticeps
and its relative, A. alfari appear to be relative
newcomers, dominating contemporary Cecropia
populations along riverine and forest edge. The
two species are closely allied to ants of other early
successional ant plants (Longino 1991b). These
ants include A. foreli Emery, which inhabits live
stems of a variety of rainforest trees, and A. longiceps
Forel, from mid-elevation Triplaris of the Costa
Rican Pacific coast. Still other representatives of
this species-group occur on Cordia alliodora. Thus,
A. ovaticeps and A. alfari may have originated
during a comparatively recent host switch onto
Cecropia. In support of this conjecture are rare
observations of apparent mistakes in colony found-
ing behavior. Queens of A. ovaticeps occasionally
attempt to enter Cecropia membrauacea by bur-
rowing into the trichilia, rather than into prostomas,
even though suitable prostomas are available in
uncolonized intemodes (D. Davidson, personal ob-
servation). The arrival of A. ovaticeps may have
driven A. australis out of riverine environments and
deeper into the forest, where it persists on a variety
of forest light-gap Cecropia species (see above;
Davidson and Fisher 1991).

Azteca australis could itself be a secondary
colonist. A member of the A. muelleri species
complex, it is likely descended from generalized

Volume 2 , Number 1 , 1993

carton-building ancestors with well-defended cen-
tral nest sites (Longino 1991a and b). Members of
this group still maintain carton masses inside the
boles of their hosts (Longino 1991a). Ants in this
species complex may have gotten their first foot-
hold on myrmecophytic Cecropici by building
external carton nests on hosts whose prior residents
(possibly Camponotus and Pcichycondyla species)
had died.

Analogously and in contemporary times, Azteca
may be invading other myrmecophytic associa-
tions. In the Manu National Park and Tambopata
Reserve of southeastern Peru, at least two carton-
building species (probably A. ulei Forel var. cordiae
Forel and A. traili [Emery] var. tococae Forel) are
residents of trichome myrmecophytes Cordia
nodosa and Tococa spp. (Appendix 1 ). Queens of
both ants initiate their colonies inside domatia
covered by protective hairs, and their incipient
colonies exhibit host-plant fidelity. Nevertheless,
larger, established colonies not only leave their
hosts regularly to forage, but build satellite nests
(often as ant-gardens) on neighboring trees. These
ants also prune trail systems through the protective
stem trichomes. On Cordia, Azteca ants occur
mainly on hosts in environments of unusually high
light intensity, and conspecific trees in the primary
forest understory are occupied by Allomerus. If we
are correct in assuming that plants with long, dense
and erect pubescence became myrmecophytes in
the context of persistent occupation by tiny and
competitively subordinate ants, then larger-bodied,
aggressive and dominant Azteca appear to have
both restricted the distribution of Allomerus, and
perhaps eliminated the former residents of Tococa.
Although Tococa is colonized occasionally by timid
Crematogaster cf. victima and a species of
Solenopsis, we have never found established colo-
nies of these ants on the Tococa of southeastern
Peru.

Identifying and Characterizing Dominants

Crematogaster and Azteca are the two genera
for which biogeographical and phylogenetic infor-
mation is most suggestive of a frequent role as
secondary colonists in species replacements among
plant-ant guilds. They are also the preeminent

competitive dominants in the arboreal ant faunas of
Africa and Asia, and New World tropics, respec-
tively. Isolated from these continents, the Austra-
lian tropics (including New Guinea and associated
islands) contains a unique set of competitive domi-
nants and relative newcomers to ant-plant symbio-
ses. Among these ants (all dolichoderines) are two
genera previously classified as Iridomyrmex
(Shattuck 1 992b), but now considered to be distinct
taxa and endemics of either the Australian
(Anonychomynna) or Oriental and Australian re-
gions (Philidris). Also included are pantropical
Technomynnex (a single species of which is appar-
ently native to the New World, Shattuck, 1992a).

Several other kinds of evidence substantiate the
inferential evidence about the relative competitive
abilities of ants involved in ant-plant symbioses.
Field experiments have demonstrated that
Crematogaster and Azteca are the principal formicid
enemies of New World Pseudomyrmex on Triplaris
(Davidson et al. 1988, see also Oliveira 1987).
Furthermore, both host-plant fidelity and pruning
of host-plant neighbors are indicative of weak com-
petitive ability (Davidson et al. 1988, 1989) and
occur with some frequency in Pseudomyrmex,
Tetraponera, Pheidole, Camponotus, and in vari-
ous myrmelachistines. In contrast, these behaviors
are atypical of Crematogaster, Anonychomynna,
Azteca, and Technomynnex (Appendix 1, column
4). Rare occurrences are limited to early succes-
sional environments where vines and competitors
are particularly threatening, as for the Azteca of
New World Cecropia, and Crematogaster of Asian
Macaranga. They can also characterize ants which
are unusually timid for their genera, as are the
Azteca exhibiting host-plant fidelity on pubescent
species of Triplaris.

Implicit in their capacity to invade
myrmecophytes previously dominated by other ants,
secondary colonists likely owe their success to
evolutionary novelties which have enhanced their
colonizing and/or competitive abilities. The genera
listed above as competitive dominants are alike in
possessing potent exocrine products which help to
convey competitive superiority in interactions with
other ants (Blum and Hermann 1978, Buschinger
andMaschwitz 1984). Structural characteristics of
waists and gasters permit workers to elevate gasters

50

Journal of Hymenoptera Research

and direct toxins toward enemy ants. The same
adaptations can be effective against potential nest
raiders, as when Crematogcister workers seal hol-
low stem nests with protruding gasters bearing
poison droplets on modified spatulate stings (Forel
1928). Many dominants are also carton-builders,
which monopolize resources in the arboreal zone
by constructing primary or ancillary nests over
Homoptera and other localized food sources such
as extrafloral nectaries.

These traits contribute to the capacity of domi-
nant ants to monopolize “promiscuous” plant re-
wards such as EFN’s and surface-feeding
Homoptera, which are either totally unprotected or
only partly secluded beneath clasping or folded
stipules of myrmecophiles. Thus in Borneo,
Crematogcister species dominate the exposed EFN’s
of most individuals of myrmecophilic
Endospermum (Euphorbiaceae), Ryparosa
(Flacourtiaceae), andMacciranga aetheadenia Airy
Shaw (D. Davidson, personal observation).
Crematogaster are also preeminent among visitors
to other myrmecophilic Malaysian Macarangaspp.
(Fiala and Maschwitz 1991). In New Guinea,
scale-tending Crematogaster are the numerically
predominant inhabitants of the stout hollow stems
of weedy Nauclea (D. Davidson, personal observa-
tion). Myrmecophiles with nectaries partly se-
cluded beneath folded or clasping stipules include
New Guinea A rchidendron (Fabaceae) and Orien-
tal Shorea (Dipterocarpaceae), both often domi-
nated by Technomyrme.x ants (D. Davidson, per-
sonal observation, Tho, fide Maschwitz and Fiala,
in press). By sealing off the folded stipules with
carton, these ants may restrict their competitors’
access to EFN. An ability to monopolize externally
located food resources may also confer a competi-
tive advantage to dominants on myrmecophytes
which produce such resources. This result would
be especially likely if evolutionary interactions of
the plants with prior ant associates had led to
increased size and/or number of EFN’s and food
bodies, or otherwise increased the rate of food
production to a level at which the plant becomes
attractive to competitive dominants requiring high
rates of resource supply.

Processes of Species Replacements

How have secondary colonists managed to re-
place primary associates with highly evolved mecha-
nisms for locating and exploiting hosts? Even very
aggressive and dominant ants may have difficulty
evicting weakly competitive ants, once the latter
have established their colonies. Thus it seems
likely that many secondary colonists first achieved
access to myrmecophytes by occupying hosts whose
usual partners were absent for one reason or an-
other. For example, like the Azteca discussed
above, some Crematogaster could have gained a
preliminary foothold on myrmecophytes by build-
ing carton nests on plants which had outlived their
ant colonies. Early stages of this scenario may be
represented in the New World associations of
Crematogaster with myrmecophytic acacia spe-
cies in second growth environments ( Janzen 1 983).
Although Crematogaster are apparently unable to
replace Pseudomyrmex on smaller acacias, they
can resist colonization by the latter species on
larger acacias which have lost their former
Pseudomyrmex colonies.

The more characteristic ant associates may be
absent for other reasons. First, by opening domatia
to feed on ant larvae, vertebrate predators of ants
may make these domatia unsuitable for continued
habitation by weakly competitive species. For
example, after swollen internodes of Cordia
alliodora are opened by woodpeckers, unspecialized
Crematogaster often move in and employ carton
baffles to seal breaks in the domatia (R. Carroll,
personal communication). Second, older domatia
are frequently abandoned by the usual residents, as
colonies move to follow new growth and produc-
tivity. In Cecropia (Davidson et al. \99\),Remijia
(Benson 1985), Leonardoxa (D. McKey, personal
observation ), Endospermum, Korthalsia, and other
genera (D. Davidson, unpublished), such aban-
doned domatia are often occupied by unspecialized
ants, which gain at least protected nest sites if not
food (Davidson and Fisher 1991, Longino 1991a).
A possible case of progressive specialization in
such ants may be seen in the unnamed
Crematogaster species which occupies Cecropia
near Genaro Herrera in Loreto, Peru (D. Davidson,
personal observation). Related to C. curvispinosa

Volume 2, Number 1, 1993

51

(J. Longino, personal communication), it is appar-
ently descended from generalized stem-nesters,
rather than from a carton-building lineage. Special-
ization on Cecropia could have been favored by
selection sharpening the host-finding abilities of
foundresses which occasionally colonized the
woody bases of forest-gap plants, and eventually
evolved to recognize Mullerian bodies as food.

Third, the typical ant associates may fail to
either colonize or to persist on hosts in inappropri-
ate habitats. Small forest light gaps are marginal for
western Amazonian Cecropia , and comparatively
low colonization rates on isolated and inconspicu-
ous gap plants appear to have provided safety for
refugees from riverbanks, as well as opportunities
for in situ colonization of this host genus. All four
little-known genera of Cecropia ants persist princi-
pally in forest light gaps. Both Camponotns balzani
and Pacltycondyla luteola colonize riverine plants,
but rarely persist there, being excluded by Azteca.
In contrast, species of Crematogaster and
Camponotns ( Pseudocolobopsis ) occur on several
light gap species at Genaro Herrera, but apparently
do not even colonize plants of riverine and forest
edge. Their relationships with Cecropia may have
evolved in situ. Alternatively, past competition
with Azteca may have led to a shift in their habitat
preferences.

Finally, colonists may also gain a foothold at the
latitudinal or elevational limits of ant-plant asso-
ciations. Latitudinally, the genus Triplaris ranges
northward into Mexico; in southwestern Chiapas
near Mapastepec, it is occupied by a variety of
apparently unspecialized species of Azteca ,
Crematogaster and Pseudomyrmex , rather than by
the more typical specialized pseudomyrmecine as-
sociates (D. Davidson, personal observation). At
least one specialized Cecropia ant, dry forest A.
coentleipennis Emery, may have evolved in situ in
Central America (Longino 1989a and b), a periph-
eral and comparatively species-poor region within
the overall distribution of Cecropia. These events
might well have resulted from independent second-
ary colonizations of a host which reached Central
America from South America in advance of its
typical ant symbionts, or which colonized habitats
unsuited to the usual associates.

Elevational segregation among plant-ants of par-
ticular hosts suggests that new colonizations might
occur at the elevational limits of species distribu-
tions. In the lowlands of Cameroon, myrmecophytic
Leonardoxa consistently house one of two closely
related myrmelachistine ants, Petalomyrmexphylax
or Aphomomyrmex afer, depending on host species
(McKey 1991). However, in submontane forests of
the Rumpi Hills (500-1700 m), where neither of
these ants occurs in association with Leonardoxa ,
the plants are inhabited by a bewildering array of
other ants, including at least two species each of
Crematogaster, Axinidris and Technomyrmex, and
one species each of Tapinoma and Leptothorax (R.
Snelling, personal communication). Some of these
ants are known not to be host-specific, and they
may be secondary colonists of a preexisting asso-
ciation of Leonardoxa with myrmelachistine ants,
although firm conclusions on directionality of this
shift must await further work. Finally, in the
Neotropics, altitudinal replacements should be com-
mon at the periphery of the Andes. Although we
know of no published data to test this prediction,
Longino (1991b) relates that the ranges of some
Azteca residents of Cecropia segregate altitudinally,
with some species occurring as high as 2000 m in
elevation.

As secondary colonists of myrmecophytes be-
come increasingly specialized for exploiting their
new hosts, selection should enhance the host-find-
ing abilities of these species. With their priority-of-
colonization eroded, primary associates may even-
tually be displaced to marginal habitats or replaced
altogether.

Taxonomic Progressions Within
Plant-Ant Lineages

Since ant-plant symbioses have been shaped by
repeated evolutionary colonizations and strong com-
petition among ants, major taxa of plant-ants might
be expected to exhibit regular taxonomic progres-
sions in species distributions and characteristics.
Similar progressions have been described for adap-
tive radiations in several well-studied animal groups,
including ants (Wilson 1959a and 1961), carabid
beetles (Erwin 1985), and birds (Ricklefs and Cox
1972; Diamond 1986). These accounts are related

52

Journal of Hymenoptera Research

in their emphasis on competition as the force driv-
ing evolutionary trajectories in animal lineages.
Wilson’s seminal exposition of the “taxon cycle” in
Melanesian ants proposes that ants invade new
geographic areas principally via marginal habitats
where competition from other ants is reduced.
From this tenuous foothold, and driven by arrivals
of new and more dominant species, they diversify
and evolve competitive strategies which eventually
enable their invasion of more species-rich forest
habitats. In apparent contrast, Erwin’s recent ac-
count of “taxon pulses” in carabid beetles proposes
that young carabid taxa first appear in productive
and central moist equatorial habitats. There, they
force the specialization and migration of older taxa
into less competitive peripheral latitudes and habi-
tats. Apparent disparities in the phrasing of Wilson’ s
and Erwin’s theories obscure theircommon ground.
Both ideas have their roots in Darlington’s (1957)
“centrifugal speciation”, whereby intense biotic
interactions drive waves of species and higher taxa
from tropical to temperate regions. Moreover,
whether species originate in new and permissive
environments, or as evolutionary novelties in
biotically restrictive environments, young species
are those with “r-selected” life histories, and gener-
alized and expanding distributions. Older, progres-
sively “K-selected” species are driven by biotic
interactions to increasing specialization and more
circumscribed distributions. There they persist by
either unique strategies for evading natural en-
emies, or by tolerance of unfavorable conditions.
Vermeij (1978) has argued cogently for similar
evolutionary trajectories in various marine inverte-
brate taxa.

The evolutionary history of plant-ants strongly
suggests similar taxonomic progressions. Three
types of evidence support such an interpretation.
First, as discussed above, taxonomic and biogeo-
graphic patterns in some ant-plant symbioses sug-
gest directionality in species replacements, and
particular taxa occupy predictable roles as victims
(e.g., Pseudomyrmecinae) and agents (e.g.,
Crematogaster and Azteca) of such replacements.
Second, and also discussed above, field experi-
ments and observations strongly support interspe-
cific competition among ants, often habitat-depen-
dent in its outcome, as the principal mechanism of

species replacements. Furthermore, roles of differ-
ent ants in postulated replacements are consistent
with their status (independently determined) in
competitive hierarchies. Third, within ant-plant
guilds, the postulated replacements of subordinate
genera, such as Pachycondyla, Plcigiolepis,
Camponotus, Pseiidomyrmex, and Tetraponera , by
dominant genera such as Crematogaster,
Technomyrmex, and Azteca, are consistent with the
historical sequence in which these taxa are repre-
sented in the fossil record (Table 2, based on
Holldobler and Wilson 1990, and see below).

The diversification of ant taxa began in earnest
no later than the beginning of the Tertiary Period
(Holldobler and Wilson 1990), and it eventually
made ants the most important natural enemies of
one another. At protected nests and feeding sites,
timid, twig-inhabiting myrmelachistines and
pseudomyrmecines, probably among the earliest
plant-ants, sought out pubescent plants or insect
borings and other cavities of live plants. But in the
background, competition was escalating. Evolu-
tionary advancements in offensive and defensive
weaponry intensified the pressures on timid and
secretive plant-ants. As discussed above, evolu-
tionary novelties and secondary colonizations ap-
pear to have arisen differentially in environments
where disturbance favored weedy species with early
and high reproductive allocation, superior coloniz-
ing abil ity , and thus priority of access to ant domatia.
Here also, high productivity (associated with high
light intensities) subsidized rapid colony growth
and the evolution of costly chemical weaponry.
Individually or in combination, these traits made
their bearers formidable enemies of existing plant-
ants, driving them into ever more restrictive spe-
cialization on one or a few hosts, into marginal
habitats, and in some cases into extinction. Eventu-
ally, many secondary colonists appear to have
partly or completely replaced the primary associ-
ates of several myrmecophyte lineages. These
secondary associates were often pressured in turn
by successive waves of newly evolved dominants.

What examples support such a scenario?
Myrmelachistine ants provide perhaps the best il-
lustration of the fate of an old group of competi-
tively subordinate ants, whose members have been
driven to suboptimal habitats, to extreme special-

Volume 2 , Number 1 , 1993

53

ization, or to extinction, by dominant ants. As
circumscribed by Holldobler and Wilson (1990),
following Wheeler (1920), this tribe is pantropical
and includes six genera, two of which are endemic
to each of the major tropical regions (the New
World, tropical Africa and the Oriental tropics). In
a recent and still incomplete analysis of generic
relationships in Formicinae, Agosti (1991) casts
doubt on the monophyly of the tribe, placing
Cladomyma in a different informal genus-group
from all the others. We follow the usual treatment
of the tribe, but acknowledge the need for further
work to resolve phylogenetic relationships of these
ants.

Myrmelachistine genera have no fossil record
(Table 2), possibly because most have long been
specialist plant-ants with restricted ecological dis-
tributions. However, they are likely to have been
widespread prior to Miocene times, since two ant
genera from a tribe (Gesomyrmecini), considered
by Wheeler (1920) to be closely related (but see
Agosti 1991), are represented in early Oligocene
Baltic amber (Holldobler and Wilson 1990). One
of these, Gesomyrmex , is represented by four extant
species of the Oriental region (Wheeler 1929a).
They share with the Oriental myrmelachistine
Cladomynna certain similarities, such as reduced
antennal segmentation (believed to be a derived
character) and worker polymorphism with major,
media, and minor workers. Furthermore, G.
kalshoveni Wheeler of Java, is recorded as nesting
in twig cavities of Artocarpus in primary forest
(Wheeler 1929b). These bits of information on an
ant genus regarded by Wheeler (1929a) as “living
fossils which have undergone no significant modi-
fication since the Early Tertiary” suggest that the
plant-ant habit may have a long evolutionary his-
tory in the Formicinae, currently regarded as hav-
ing diverged very early from the basal lineage of the
Formicidae (Holldobler and Wilson 1990).

In all parts of their pantropical distribution,
myrmelachistines appear to have experienced eco-
logical contraction. Although no phylogeny is
available for the New World genus Mynnelachista ,
interspecific patterns in its distribution and ecology
reveal the likely imprint of past competition.
Mynnelachista are often conspicuous leaf foragers
in montane forests of Central and South America,

where dominant Crematogaster and Azteca ants
are largely missing (J. Longino, personal commu-
nication). In sharp contrast, congeners of tropical
lowlands are stem-nesters with a relatively incon-
spicuous presence on leaf surfaces. Among resi-
dents of Costa Rican Ocotea, workers of a
Mynnelachista plant-ant at 500-700 m elevation (at
Rara Avis) do not attack vines (B. Fisher, personal
communication), though those of a congener at 50
m in nearby La Selva Biological Station do prune
(D. Davidson, personal observation). Finally, in
western Amazonia, perhaps the center of neotropical
ant diversity (Wilson 1987), Mynnelachista resi-
dents of Duroia hirsuta and Cordia nodosa appear
to protect themselves not only by pruning vegeta-
tion other than potential host plants, and by main-
taining extensive clearings (“supay chacras”), but
by effectively hiding from larger-bodied ants amid
the dense stem hairs of these two hosts. (Morawetz
et al. [ 1 992] argue that creation of similar clearings
by a Mynnelachista species on Tococa is not a
product of past competition. However, this asser-
tion is based strictly on the probably valid assump-
tion that clearings enhance the light environment
and productivity of host plants; it did not stem from
any direct test for the effects of competition from
other ants [see, e.g., Davidson et al. 1988]). Over-
all, the pattern reveals that increasing specializa-
tion for resisting dominant ants may have been
required for persistence in highly competitive and
diverse lowland rainforest faunas.

The evolutionary fortunes of myrmelachistines
also appear to have declined in the Old World
tropics. In Africa, they are represented by only two
monotypic genera ( Petalomynnex and
Aphomomynnex). The former is restricted to a
single host species and confined to a very small area
of Lower Guinea coastal forest. Both species are
plant-ants, though interestingly, neither prunes nor
inhabits pubescent myrmecophytes. Cladomynna
is one of two myrmelachistine genera known from
Asia (with the status of Pseudaphomomyrmex re-
maining uncertain), and all five described species
are specialized plant-ants (Agosti 1991). Some of
their hosts (e.g., Saraca) are shared with
Crematogaster , suggesting the potential for com-
petitive interactions with this group of dominant
ants. Furthermore, patterns of host association indi-

54

Journal of Hymenoptera Research

cate that Crematogaster may have replaced
Cladomyrma in some systems. Thus Cladomynna
persists on Asian Neonauclea, but Crematogaster
dominates closely related Myrmeconauclea. Too
little is known of phylogenetic relationships among
representatives of any of these lineages to draw
firm conclusions.

Pseudomyrmecines appear to be another rela-
tively old group in which the plant-ant habit may be
ancient, and in which competitively subordinate
plant-ants have been restricted or replaced by more
recently evolved, competitive dominants.
Tetraponera first appears in fossil deposits in the
early Oligocene and Pseudoinynnex in the Oli-
gocene (Table 2). The monotypic Myrcidris, a
plant-ant whose specializations indicate a long his-
tory of association with plants, may be a relict that
is the sister group to all other pseudomyrmecines,
though other interpretations are possible (Ward
1990). As discussed above, plant-ants of this rela-
tively old subfamily are among the most frequent
apparent victims of the expansion of younger groups
such as Crematogaster and Azteca.

Other groups of competitively subordinate ants
tor which there is circumstantial evidence of re-
placement by more recently evolved dominants
also occur relatively early in the fossil record.
These include Pachycondyla, Camponotus, and
Plagiolepis , all of which appear in the early Oli-
gocene. Cecropia specialists derived from widely
distributed Pachycondyla villosa and P. unidentata
(J. Longino, personal communication) are prob-
ably more recent secondary colonists, inhabiting
mainly older and woody stems abandoned by other
ants. At present, no evidence indicates that these
are replacing former inhabitants. Allomerus, an-
other genus being pressured by contemporary domi-
nants, has no fossil record, perhaps because all of
these ants have been plant-ants with highly re-
stricted distributions.

In contrast to these weakly competitive groups,
genera implicated as dominant ants and^ secondary
or tertiary colonists of existing associations appear
to be more recent arrivals. The first fossil records
of Crematogaster and Technomyrmex are in the
Miocene, and Azteca appears in the early Miocene
(Table 2).

Taxonomic Progressions and Intercontinental
Comparisons of Ant-Plant Symbioses

If taxonomic progressions such as those postu-
lated above play major roles in transforming ant-
plant symbioses over evolutionary time, then long-
term evolutionary history assumes an added di-
mension as an important factor shaping interconti-
nental differences in the nature of ant-plant sym-
bioses. Contemporary patterns will reflect the
point to which taxonomic progressions in plant-
ants have proceeded in a region. The location of
this point should depend on the ages of regional
mesic-forest communities (to which most ant-plant
symbioses are restricted), the traits of the particular
dominant and subordinate ants evolved there dur-
ing this period, and the degree to which the region
is isolated from the products of taxonomic progres-
sions begun elsewhere.

West Gondwanaland, today represented by its
derivative continents Africa and South America,
has been considered the cradle of the angiosperms
(Raven and Axelrod 1974). Mesic tropical forest
and its typical constituents, including plant-ants,
have had a long history on both these continents. In
Africa, for example, despite climatic vicissitudes
and shifts in continental position, a large area of
lowland tropical rain forest has persisted unbroken
since the Late Cretaceous-Paleocene (75-55 my
B.P.) up to the present (Axelrod and Raven 1978).
That taxonomic progressions in Africa and South
America began with similar starting material, and
have continued for about the same amount of time,
may account for many of the striking similarities in
ant-plant symbioses of these two regions (McKey
and Davidson, in press). Interestingly, these two
continents share old, competitively subordinate ant
groups like myrmelachistines and pseudo-
myrmecines. Although these taxa would respond
in analogous ways to the later onslaught of domi-
nants, the dominants are derived from different
genera on the two continents. Whereas in the
Neotropics, the preeminent competitive dominants
consist of endemic Azteca, Crematogaster domi-
nate in the Old World, where they are much more
prevalent than in the American tropics (Appendix
1).

Volume 2, Number 1 , 1993

55

During virtually all the Tertiary, South America
was an island continent (Barron et al., 1 98 1 , Gentry
1982). Perhaps the later appearing dominants,
Crematogaster and Azteca , evolved long after di-
rect exchange between the two continents (via
overland connections or island filter bridges) be-
came impossible. Evidence suggests that
Crematogaster could be an Old World genus which
arrived relatively late in the New World, possibly
as part of a widespread tropical Laurasian biota,
elements of which could have reached the Neotropics
via North America. First recorded in Sicilian
amber in the Miocene, the genus is represented in
Dominican amber (late Miocene), and might con-
ceivably have invaded South America via Panama,
a connection in place since the Pliocene (Keigwin
1978, Barron et al. 1981, Marshall et al. 1982).
Moreover, species richness of Crematogaster is
greater in the African and Oriental tropics than in
the Neotropics (Brown 1973), and the genus has
evolved numerous specialized plant-ants in the
former two regions, but only two such described
species in the American tropics. From our sum-
mary in Appendix 1, relationships involving
Crematogaster account for only 7.6 % of all 66
symbiotic ant-plant relationships listed for the
Neotropics, but 39.5 % of 43 associations and 27.3
% of 33 relationships in Africa (including Mala-
gasy) and the Oriental tropics, respectively. Based
on analyses at the generic level, our calculations fail
to take into account the substantial radiations of
species within the genus Crematogaster on Asian
Macaranga (Appendix 1 ), as well as the nine spe-
cies of Crematogaster occurring on African
Mnsanga (though probably none is a specialized
plant-ant). No parallel radiations occur in the
American tropics.

During the Tertiary, while the South American
biota was evolving in isolation, there were repeated
opportunities for biotic exchange between tropical
Africa and tropical Laurasia. The latter region has
long harbored mesic tropical forests, though opin-
ions vary on whether these forests are as ancient as
those of West Gondwanaland (Raven and Axelrod
1 974). At the very least, the Oriental tropics were
an area of moist, equable climate relatively re-
moved from the major vicissitudes of Neogene and
later climatic change (Raven and Axelrod 1974).

Biotic connections of tropical alliances, at least
through the early Tertiary, may account for simi-
larities in taxonomic composition of both subordi-
nate and dominant plant-ants of the African and
Oriental regions (e.g., Tetraponera as well as
Crematogaster and Oecophylla). They may also
help to explain some possible cases of common
ancestry among ant-plant associations of the Afri-
can and Oriental tropics (McKey and Davidson, in
press).

Of the major tropical regions, the Australian
tropics (northern Australia, New Guinea, and asso-
ciated islands) are outstanding for the geologic
youth of their tropical mesic-forest environments.
By the Paleocene, Australia was connected with the
rest of the world only by a cool-temperate pathway
to South America via Antarctica (Raven and Axelrod
1 974). At the start of its northward movement 45-
49 my B.P., what is now tropical northern Australia
was all well south of the Tropic of Capricorn, and
was still 1 0 degrees south of its present position by
the Miocene, when direct migration from the Asian
tropics first became possible (Axelrod and Raven
1972). As for New Guinea, neither it nor its
principal antecedents existed prior to about 40 my
B.P. Only by the Miocene did it lie close enough to
the proto-Indonesian arc to begin receiving large
numbers of immigrants from tropical Asia. (How-
ever, as vertebrate distributions illustrate, such
migration was never directly overland [Axelrod
and Raven 1982]). Thus tropical northern Australia
and mesic-forest portions of New Guinea have
been populated to a large degree by taxa derived
from the Asian tropics via intervening islands (Wil-
son 1961; Raven and Axelrod 1 974). Nevertheless,
the contemporary distributions of at least some
plant-ants (e.g., Anonychomynna, see Shattuck,
1 992b) reveal an almost certain origin in Australasia.

Tropical forests of northern Australia and New
Guinea provide uniquely little evidence for re-
placement of older and competitively subordinate
ant genera by contemporary dominants. The ori-
gins of tropical rain forests in the Australian region
have apparently been too recent to have allowed
significant radiations of specialized plant-ants in
more ancient and weakly competitive ant genera
prior to the arrival and expansion of the dominants.
If so, this could help to explain why the fraction of

56

Journal of Hymenoptera Research

ant-plants obviously specialized as myrmecophytes
is so low in the Australian region (column 6 in
Appendix 1 ). Compared to 5 1 .2 % of 39 Neotropi-
cal ant-plant genera, 59.4 % of 32 African genera,
and 4 1 .4 % of 29 Oriental myrmecophyte genera,
only 10.7 % of 28 such genera in the Australian
region have conspicuous specializations to attract
ants. In the last of these areas, only Endospermitm,
Ccmthium and Calamus have convincingly ant-
attractive traits (Appendix 1). Present day plant-
ants of this region consist principally of dominant
species of Anonychomyrma (formerly included in
Iridomyrmex, Shattuck 1992b), Teclinomyrmex and
Crematogaster, as well as Philidris on epiphytic
myrmecophytes (Shattuck 1992b). These ants oc-
cupy only a small number of variously preadapted
host genera, where they maintain scale insects at
remarkably high biomass, possibly limited by stem
volume. Consistent with their status as dominants,
they do not exhibit host fidelity in foraging. Neither
pruning of host-plant neighbors, nor hiding among
dense trichomes is required for persistence of such
capable competitors. Two associations with weakly
competitive ant genera may also be comparatively
recent in origin. The Camponotus of Endospermum
obtain their protein not from specially evolved
plant structures, nor from protected sources within
the stem (e.g., homoptera or the heteroplasias of,
e.g., Vitex ), but through a form of parasitism of
external stem walls, i.e., the induction of
heteroplasias from cambium (D. Davidson, per-
sonal observation). Moreover, at least one pro-
posed myrmecophyte in this genus often occurs
without its ants (Airy-Shaw 1980). Similarly, Ward
(1991) notes that the unnamed Tetraponera tend-
ing coccids in terminal branches of Cupaniopsis
has a much narrower geographic range than does its
host, and that the symbiosis is apparently young.

In attempting to explain intercontinental differ-
ences in diversity, it will be extremely difficult to
distinguish the relative importances of two major
historical factors. These are regional differences in
1) the condensation of diversity through competi-
tion; and 2) the magnification of diversity, as af-
fected by habitat diversity and its effects on rates of
evolutionary host shifts and de novo evolutionary
colonizations (see above and McKey and Davidson,
in press.).

CONCLUSIONS

Similar selection pressures acting on correspond-
ingly preadapted ants and plants have produced
strikingly parallel and convergent evolution in the
symbiotic ant-plant relationships of different tropi-
cal regions. Although current concepts of ant-plant
coevolution focus on the pairwise interaction be-
tween ant and host plant, these alone cannot ac-
count for the patterns we observe. Even in relation-
ships where pairwise interactions are undoubtedly
strong, multispecies interactions appear to have
determined many features of present-day symbio-
ses. The most important force driving the evolu-
tionary biology of ant-plant symbioses is interspe-
cific competition among arboricolous ants. Plants
differ in the kinds of resources which they offer to
ants, in the rates at which they supply these re-
sources, and in traits which influence the relative
competitive abilities of foraging and nesting ants.
As in other communities structured by competition,
plant-ants sort out across plants in ways that are
predictable from their particular resource require-
ments and competitive abilities and the spectrum of
available resources (see also Bristow 1991). In the
American, African and Asian tropics, competi-
tively dominant ants are associated with the most
light-demanding and fast-growing hosts, which
supply resources at the rates required to fuel rapid
colony growth, interspecific aggression and other
traits required for dominance. In contrast, competi-
tively subordinate ants are restricted to plants which
supply resources at rates too low to support domi-
nant ants, or to those from which dominant ants can
be excluded by long, dense plant hairs, pruning of
neighboring vegetation, or by other ant and plant
traits which favor competitively subordinate spe-
cies. Competitive interactions among ants deter-
mine whether patterns of ant-plant association are
sufficiently predictable for strong interactions to
shape the evolution of ants and plants. When
competitive interactions in plant-ant guilds result
in constancy in the pairing of particular ants and
plants, reciprocal evolutionary interactions may
occasionally give rise to pairwise coevolution.

Parallel and convergent selection pressures acted
on similar biological material on different tropical
land masses. In American, African, Asian and

Volume 2 , Number 1 , 1993

57

Australian regions, the same important
preadaptations facilitated evolution of the plant-ant
habit in several lineages of arboricolous ants. Fore-
most among these traits were the habit of tending
Coccoidea, and the differential competitive abili-
ties determined by generically typical offensive
and defensive weaponry, or by inherent colony
growth rates and other life-history attributes. Like-
wise, similar sets of plant traits facilitated the
evolution of myrmecophytes on different conti-
nents. Structures evolved independently of ant-
related selective pressures were co-opted repeat-
edly as myrmecophytic traits in plant lineages that
eventually produced ant-plants. These traits in-
cluded both the long, dense hairs typical of many
myrmecophyte stems and domatia, and stems
strongly thickened as support structures for large
leaves, and available as nest sites for opportunistic
ants. These similarities in starting material have
rendered even more pronounced the striking paral-
lel and convergent evolution of ant-plant symbio-
ses in the New World and Old World tropics.

Diversity of both myrmecophytes and their at-
tendant ants appears to accumulate mainly across
habitats, rather than biogeographical regions
(McKey and Davidson, in press). Among ant-
plants, evolutionary diversification across habitat
boundaries often appears to reflect the conflicting
selection pressures imposed by different plant re-
source environments. Like other tropical plants
(McKey etal. 1978, Coley 1983), myrmecophytes
have responded evolutionarily to particular resource
regimes by altering their relative investments in
defense versus growth and, perhaps, their relative
allocation of different kinds of resources to defen-
sive function (Davidson and Fisher 1991, Folgarait
and Davidson 1992). In turn, ecological and evolu-
tionary responses of plants to different resource
environments determine the quantity and quality of
resource supply to ants. On the whole, then, both
partners in ant-plant associations may be more
sensitive to habitat than to taxonomic differences
among symbiotic partners.

Strong competition among mutualists has been
proposed as a major factor driving the evolution of
specialization in mutualisms (Law and Koptur
1 986), and it could help to account for the origins of
many specialized ant-plant symbioses. Neverthe-

less, where sufficiently well-studied, phylogenies
of plant-ants, together with host distributions of
these ants, suggest that pairwise coevolution and
cospeciation have been rare. Rather than simple,
pairwise ant-plant systems, guilds of interacting
ants and plants seem to be the most frequent arena
of ant-plant evolutionary interaction. Perhaps as a
consequence, plant-switching and secondary colo-
nization (rather than cospeciation or some other
form of association by descent) may have been the
usual processes by which these mutualisms diver-
sified. Repeated colonization of myrmecophyte
taxa has occurred as unspecialized ants have ex-
ploited preexisting mutualisms and specialized
plant-ants have switched hosts. Habitat-depen-
dence in the effect of associations on fitness of the
participants seems to have been the principal force
leading to the evolution of new associations. The
motor driving such evolutionary opportunities was
likely the climatically induced range expansion
that placed ants or plants into habitats sufficiently
novel to change selection regimes, and to increase
encounters with new associates (McKey and
Davidson, in press).

Regardless of how species originate, a complex
mosaic of habitats should help to maintain higher
local diversity, with greater species richness of
myrmecophytes and/or specialist plant-ants, and a
greater number of ant/plant combinations. Since
the potential for evolution of new associations via
host shifts and secondary colonization depends in
part on the sizes of locally interacting ant and plant
guilds, high local diversity may lead to higher rates
of species origination. Thus, independently of
distributional changes driven by varying climates,
beta-diversity is likely to have enhanced alpha-
diversity.

As summarized here, the determinants of diver-
sity of plants and ants in these symbiotic mutualisms
will likely generalize to other components of tropi-
cal floras and faunas. In particular, we expect that
diversification of tropical plants has often involved
evolutionary adjustments in the amounts and kinds
of defenses, in response to habitat differences in
absolute and relative availabilities of essential re-
sources. Consequently, habitat mosaics related to
edaphic factors and incident solar radiation should
often determine mosaics in the primary productiv-

58

Journal of Hymenoptera Research

ity available to consumer organisms. Habitat spe-
cialization to different productivity regimes has
likely been important to both the generation and
maintenance of diversity in many tropical con-
sumer guilds whose member species have strongly
overlapping resource requirements (cf., Terborgh
1983 for primates, and S. Robinson and J. Terborgh,
personal communication, for birds).

Habitat specialization may frequently also rep-
resent the intermediate and final stages of a taxon
pulse, in which new, opportunistic, abundant, and
widespread species are driven to progressively
greater specialization, finer niche differentiation,
diminished distribution and abundance, and per-
haps even eventual extinction (Wilson 1959a and
1961, Ashton 1969, Erwin 1985, Diamond 1986).
If taxon cycles or pulses are general features of
animal and plant lineages, they might aid in ex-
plaining patterns in the relative abundance distribu-
tions of taxa within higher taxa. Dial and Marzluff
(1989) have discussed the frequency of “hollow
curve distributions”, or the overdominance of par-
ticular minor taxa within major taxa (subunits/
unit). Thus, the degree of dominance of most
dominant taxa is greater than that predicted by a
variety of null models based on Poisson processes,
random cladogenesis, and simultaneous or sequen-
tial resource subdivision, and it is compounded at
lower levels of the taxonomic hierarchy. Taxon
pulses might regularly give rise to such patterns if
the enumeration of taxa at successively lower lev-
els of the taxonomic hierarchy (where taxa are more
numerous) were more likely to pick up compara-
tively rare groups which had recently acquired
evolutionary novelties, and which represented in-
termediate stages of a taxon pulse.

Our analyses of the evolutionary dynamics of
ant-plant symbioses here and elsewhere (McKey
and Davidson, in press) lead us to propose new
hypotheses to explain differences in the diversity of
ant-plants and plant-ants across different tropical
regions. Such disparities are quite pronounced
between the American and African tropical re-
gions, where ant-plant symbioses are best under-
stood. Previous explanations for differences in the
biodiversity of species-rich Neotropical and depau-
perate African rain forests have emphasized the
contrasting climatic histories of these two regions.

Focusing on range contractions during periods of
unfavorable climate, these explanations attribute
Africa’s lower diversity to greater extinction dur-
ing the Pleistocene, as Africa’s climate became
drier, and refugia were fewer than in Amazonia
(Raven and Axelrod 1974). We propose that differ-
ences between the two regions in the rates of
species origination may be at least as important as
extinction rates. The relatively stable geological
history of most of Africa, including the rainforest
zone, has created a landscape with relatively little
elevational relief (hence few sharp spatial contrasts
in temperature and rainfall), comparatively little
edaphic variation, and relatively infrequent and
spatially limited fluvial disturbance. In contrast,
the Andean orogeny, and subandean tectonic activ-
ity, have helped to create a landscape of great
elevational, climatic, and edaphic complexity, es-
pecially in western Amazonia. This has resulted in
a complex and dynamic mosaic of habitats.
Colinvaux (in press) suggests that species origina-
tion usually takes place when ranges are re-expand-
ing during periods of climatic amelioration. If this
is so, then in the Neotropics, especially in western
Amazonia, range expansion would be much more
likely than in the African forest zone to place ants
and plants into novel habitats, leading to speciation,
the formation of new associations, or both.

Although poorly known by comparison, Asian
rain forests occur in regions (especially Borneo)
where topography is substantially more variable
than that of tropical Africa. Aided by forest frag-
mentation on numerous island land masses, this
topography has contributed to the diversification of
several myrmecophyte lineages here.
Myrmecophyte diversity in Asia appears to be
intermediate between that of the African and Ameri-
can tropics. On the other hand, both regions of the
Old World tropics may have had the generic diver-
sity of their plant-ant faunas condensed, relative to
that of the New World, by the comparatively early
arrival of a new wave of competitively dominant
Cremalogcister. The relatively recent origin of rain
forests in the Australian tropics (including New
Guinea and associated islands) appears to have
limited the diversification of myrmecophytes in all
but the epiphytic Hydnophy tinae ( Jebb 1991, Huxley
and Jebb 1991). In addition, the elaboration of

Volume 2, Number 1, 1993

59

myrmecophytic traits, which have evolved so fre-
quently elsewhere in associations with weakly com-
petitive ants, may have been limited in Australia by
the cooccurrence of contemporary dominant and
subordinate ants at the time when rain forests were
evolving.

ACKNOWLEDGEMENTS

This work was supported by the NSF grants RI1-
8310359 and BSR-9003079, the Guggenheim Founda-
tion, the Christensen Research Institute and the Univer-
sity of Utah’ s Faculty Research Committee (to Davidson)
and the Swiss Natural History Society, the National
Geographic Society, and the University of Miami (to
McKey). We thank R. Snelling for numerous ant iden-
tifications, graciously fit into an overburdened schedule,
and for helpful conversations about ants. J. Longino and
P. Ward made invaluable comments on several previous
drafts of the manuscript, and they and R. Snelling made
available many of their unpublished observations. Any
mistakes remaining are our own. We are also very
grateful to M. Hossaert for facilitating our extensive E-
Mail correspondence. The manuscript benefitted im-
measurably from consultations over recent years with
myrmecologists and botanists specializing in particular
taxa or floras. These include: C. C. Berg, W. L. Brown,
Jr., W. Burger, J. Dransfield, A. H. Gentry, W. Judd, J.
Longino, J. Miller, T. Musthak Ali, S. Renner, S. O.
Shattuck, J. C. Solomon, C. M. Taylor, H. van der Werff,
P. S. Ward, and J. L. Zarucchi. Finally, since ant-plant
symbioses are being drastically altered by habitat de-
struction around the world, we are particularly grateful
for the natural parks and reserves that have permitted us
and others to study these interactions in their pristine
form.

LITERATURE CITED

Addicott, J. 1 985. Competition in mutualistic systems,
pp. 217-247. In Boucher, D. H., ed.. The Biology of
Mutualisms. Oxford University Press, Oxford.
Agosti, D. 1991. Revision of the Oriental ant genus
Cladomyrma , with an outline of the higher classifica-
tion of the Formicinae (Hymenoptera: Formicidae).
Systematic Entomology 16: 293-310.

Airy-Shaw, H. K. 1980. The Euphorbiaceae of New
Guinea. Kew Bulletin , Additional Series 8: 1-243.
Anderson, A. N. 1991. Parallels between ants and
plants: implications for community ecology, pp. 539-
558. In Huxley, C. R. and Cutler, D. F., eds. Ant -

Plant Interactions. Oxford University Press, Ox-
ford.

Anderson, M. 1991. An ant-aphid interaction: Formica
fusca and Aphthargelia symphoricarpi on Mount St.
Helens. American Midland Naturalist 125: 29-36.

Ashton, P. S. 1969. Speciation among tropical forest
trees: some deductions in the light of recent evidence.
Biological Journal of the Linnean Society 1 : 1 55-
196.

Attygalle, A. B. and E. D. Morgan. 1984. Chemicals
from the glands of ants. Chemical Society Review 13:
245-278.

Axelrod, D. l.and P. H. Raven. 1972. Paleobiogeography
and origin of the New Guinea flora, pp. 919-941. In
Gressitt, J.L., ed., Biogeography and Ecology of New
Guinea. Dr. W. Junk Publishers, The Hague.

Axelrod, D. I. and P. H. Raven. 1978. Late Cretaceous
and Tertiary vegetation history of Africa, pp. 77- J 30.
In Werger, M. J. A., ed., Biogeography and Ecology
of Southern Africa, Volume I. Dr. W. Junk Publish-
ers, The Hague.

Bailey, 1. W. 1992. The anatomy of certain plants from
the Belgian Congo with special reference to
myrmecophytism. Bulletin of the American Museum
of Natural History 45: 585-621 .

Barron, E. J., C. G. A. Harrison, J. L. Sloan 11, and W. W.
Hay. 1981. Paleogeography, 1 80 million years ago
to the present. Eclogae Geologicae Helveticae 74:
443-470.

Beattie, A. J. 1985. The Evolutionary Ecology of Ant-
Plant Mutualisms. Cambridge University Press,
Cambridge.

Beccari, O. 1884. Piante ospitatrici, ossia piante
formicarie della Malesia e della Papuasia. Malesia
(Genoa), 2: 1-340.

Becerra, Judith X. I. and D.L. Venable. 1989. Extrafloral
nectaries: a defense against ant-Homoptera
mutualisms? Oikos 55: 276-279.

Bengo, M.D. and J. Maley. 1991. Analyses des flux
polliniques sur la marge sud du Golfe de Guinee
depuis 135 000 ans. Comptes rendues de l 'Academie
des Sciences de Paris Serie 7/313: 843-849.

Benson, W. W. 1985. Amazon ant-plants, pp. 239-266.
In Prance, G. T. and T. E. Lovejoy, eds., Amazonia.
Pergamon Press, New York.

Bequaert, J. 1922. Ants of the American Museum
Congo Expedition. A contribution to the myrmecol-
ogy of Africa. IV. Ants in their diverse relations to the
plant world. Bulletin of the American Museum of
Natural History' 45: 333-583.

Berg, C. C. 1978. Especies de Cecropia da Amazonia
Brasileira. Acta Amazon ica 8: 149-182.

60

Journal of Hymenoptera Research

Berg, C. C., R. W. A. P. Akkermans, and E. C. H. van
Heusden. 1990. Cecropiaceae: Coussapoa and
Pourouma with an introduction to the family. Flora
Neotropica, Monograph 51. New York Botanical
Garden, New York.

Bloom, A. J., F. S. Chapin, III, and H. A. Mooney. 1985.
Resource limitation in plants — an economic anal-
ogy. Annual Review of Ecology and Systematics 16:
363-392.

Blum, M. S. and H. R. Hermann. 1978. Venoms and
venom apparatusesof the Formicidae: Myrmeciinae,
Ponerinae, Dorylinae, Pseudomyrmecinae,
Myrmicinae, and Formicinae, pp. 80 1 -869. In Bettini,
S.,ed ., Arthropod Venoms. Springer- Verlag, Berlin.

Bower, F. 1 887. On Humboldtia laurifolia, Vahl, as a
myrmekophilous plant. Proceedings of die Philo-
sophical Society of Glasgow 18: 320-326.

Brandbyge, J. 1986. A revision of the genus Triplaris
(Polygonaceae). Nordic Journal of Botany 6: 545-
570.

Brandbyge, J. 1 990. Woody Polygonaceae from Brazil:
new species and a new interpretation. Nordic Journal
of Botany 10:155-160.

Breteler, F. J. 1986. The African Dichapetalaceae. A
Taxonomic Revision, IX. Agricultural University of
Wageningen Papers 86-3, 74 pp.

Bristow, C. M. 1991. Why are so few aphids ant-
tended?, pp. 104-1 19. In Huxley, C. R. and D. F.
Cutler, eds. Ant - Plant Interactions. Oxford Univer-
sity Press, Oxford.

Bronstein, J. L. 1991. Mutualism studies and the study
of mutualism. Bulletin of the Ecological Society of
America 72: 6-8.

Bronstein, J. L. 1992. Seed predators as mutualists:
ecology and evolution of the fig / pollinator interac-
tion, pp. 1-44. In Bemays, E. A., ed., Insect-Plant
Interactions , Volume IV. CRC Press, Boca Raton,
Florida.

Brown, W. L., Jr. 1962. The neotropical species of the
ant genus Strumigenys Fr. Smith: synopsis and keys
to the species. Psyche 69: 238-267.

Brown, W.L., Jr. 1973. A comparison of the Hylean and
Congo- West African rain forest ant faunas, pp. 161-
186. In Meggers, B. J., E. S. Ayensu and W. D.
Duckworth, eds. Tropical Forest Ecosystems in
Africa and South America. Smithsonian Institution
Press, Washington, D. C.

Buckley, R. C. 1987. Interactions involving plants,
homoptera, and ants. Annual Review of Ecology and
Systematics 18: 1 11-35.

Burger, W.C. 1971. Piperaceae. Fieldiana, Botany 35:
5-227.

Burger, W. C. and H. van der Werff. 1990. Flora
Costaricensis, Family # 80 Lauraceae. Fieldiana,
Botany 23: 1-138.

Buschinger, A. and U. Maschwitz. 1984. Defensive
behavior and defensive mechanisms in ants, pp. 95-
1 50. In Hermann, H. R. ed.. Defense Mechanisms in
Social Insects. Praeger Scientic, New York.

Campbell, D. G., P. M. Richardson and A. Rosas, Jr.
1989. Field screening for allelopathy in tropical
forest trees, particularly Duroia hirsuta, in the Bra-
zilian Amazon. Biochemical Systematics and Ecol-
ogy 17: 403-407.

Carroll, C. R. 1979. A comparative study of two ant
faunas: the stem-nesting ant communities of Liberia,
West Africa and Costa Rica, Central America. Ameri-
can Naturalist 1 13: 551-561.

Carroll, C. R. 1983. Azteca ( Hormiga azteca, Azteca
ants, Cecropia ants), pp. 691-693. In Janzen, D. H.
ed., Costa Rican Natural History 691-693. Univer-
sity of Chicago Press, Chicago.

Carroll, C. R. and D. H. Janzen. 1973. The ecology of
foraging by ants. Annual Review of Ecology and
Systematics 4: 231-258.

Coley, P. D. 1 983. Intraspecific variation in herbivory
on two tropical tree species. Ecology 64: 426-433.

Colinvaux, P. In press. Pleistocene biogeography and
diversity in tropical forest of South America and
Africa. In Goldblatt, P., ed.. Biological Relation-
ships between Africa and South America. Yale
University Press, New Haven.

Connell, J. H. and R. O. Slatyer. 1977. Mechanisms of
succession in natural communities and their role in
community stability and organization. American
Naturalist 111: 1119-1144.

Coode, M.J.E. 1981. Elaeocarpaceae, pp. 38-185. In
Henty, E.E., ed., Handbooks of the Flora of Papua
New Guinea, 2. Melbourne Press, Melbourne.

CSIRO. 1970. The bisects of Australia. Melbourne
University Press, Melbourne, 1029 pp.

Cushman, J. H. and J. F. Addicott. 1991. Conditional
interactions in ant-plant-herbivore mutualisms, pp.
92-103. In Huxley, C. R. and D. F. Cutler, eds.. Ant
- Plant Interactions. Oxford University Press, Ox-
ford.

Darlington, P. J., Jr. 1957. Zoogeography: The Geo-
graphical Distribution of Animals. John Wiley and
Sons, New York.

Davidson, D. W. 1988. Ecological studies of neotropical
ant-gardens. Ecology 69: 1138-1152.

Davidson, D. W. and W. W. Epstein. 1989. Epiphytic
associations with ants, pp. 200-233. In Liittge, U.
(Ed.), Vascular Plants as Epiphytes. Springer-Verlag,
Berlin.

Volume 2, Number 1, 1993

61

Davidson, D.W. and B. L. Fisher. 1991. Symbiosis of
ants with Cecropia as a function of the light regime,
pp. 289-309. In Huxley, C. R. and D. K. Cutler eds..
Ant - Plant Interactions. Oxford University Press,
Oxford.

Davidson, D. W„ R. B. Foster, R. R. Snelling, and P. W.
Lozada. 1991. Variable composition of some tropi-
cal ant-plant symbioses, pp. 1 45- 1 62. In Price, P. W.,
T. M. Lewinsohn, G. W. Fernandes and W. W.
Benson, eds., Plant-Animal Interactions: Evolution-
ary Ecology in Tropical and Temperate Regions.
John Wiley & Sons, Inc., New York.

Davidson, D. W„ J. T. Longino, and R. R. Snelling.
1988. Pruning of host plant neighbors by ants: an
experimental approach. Ecology 69: 801-808.

Davidson, D. W., R. R. Snelling and J. T. Longino. 1 989.
Competition among ants for myrmecophytes and the
significance of plant trichomes. Biotropica 21 : 64-
73.

de Candolle, C. 1916. Piperaceae Philippinenses novae
vel nuper repertae. Philippine Journal of Science
(Botany) 1 1 : 207-225.

Dejean. A. and C. Djieto. 1990. The choice of the
nesting site by Oecophylla longinoda : roles of im-
printing and selective attraction to plants, pp. 85-86.
In Veeresh, G. K„ B. Mallik and C. A. Viraktamath,
eds.. Social Insects and the Environment. Proceed-
ings, 1 1 th International Congress of the International
Union for the Study of Social Insects. Oxford Uni-
versity Press, New Delhi.

Dejean, A., C. Djieto and S. Ngnokam. 1992. Les
relations plantes-fourmis: nouvel apercu. Memoires
de la Societe royale beige d’Entomologie 35: 563-
567.

Dejean, A., R. Mony, S. Ngokam and C. Djieto. 1990.
Arboreal nesting in various African ants, p. 659. In
Veeresh, G. K., B. Mallik and C. A. Viraktamath,
eds.. Social Insects and the Environment. Proceed-
ings, 1 1 th International Congress of the International
Union for the Study of Social Insects. Oxford Uni-
versity Press, New Delhi..

de Wit, H. C. D. 1942. Conspectus of the genus
Archidendron F. von Mueller (Legum.). Bulletin of
the Botanical Garden of Buitenzorg 17: 256-272.

Dial, K. P. and Marzluff, J. M. 1989. Nonrandom
diversification within taxonomic assemblages. Sys-
tematic Zoology 38: 26-37.

Diamond, J. 1986. Evolution of ecological segregation
in the New Guinea montane avifauna, pp. 98- 125. In
Diamond, J. and T. J. Case, eds.. Community Ecol-
ogy. Harper and Row, New York.

Ding Hou. 1978. Anacardiaceae, pp. 395-548. In van
Steenis, C. G. G. J. and W. J. J. O. de Wilde, eds..

Flora Malesiana Series I, Volume 8 . Kluwer,
Dordrecht.

Dorow, W.H.O.andU.Maschwitz. 1990. TheArachne-
group of Polyrachis (Formicidae, Formicinae):
weaver ants cultivating homoptera on bamboo.
Insectes Sociaux 37: 73-89.

Dransfteld. J. 1978. New records of rattans (Palmae:
Lepidocaryoideae ) for Peninsular Malaysia. Malayan
Nature Journal 30: 523-531 .

Dransfield, J. 1981. A synopsis of the genus Korthalsia
(Palmae: Lepidocaryoideae). Kew Bulletin 36: 163-
194.

Dransfield, J. 1984. The Rattans of Sabah. Forest
Department, Sabah, 1 82 pp.

Duviard, D. and P. Segeren. 1 974. La colonisation d’un
myrmecophyte, le parasolier, par Crematogaste r spp.
(Myrmicinae) en Cote-dTvoire forestiere. Insectes
Sociaux 21: 191-212.

Eisner, T. 1957. A comparative morphological study of
the proventriculus of ants (Hymenoptera:
Formicidae). Bulletin of the Museum of Compara-
tive Zoology, Harvard 1 16(8): 439-490.

Erwin, T. L. 1985. The taxon pulse: a general pattern
of lineage radiation and extinction among carabid
beetles, pp. 437-472. In Ball, G. E., ed.. Taxonomy,
Phytogeny, and Zoogeography of Beetles and Ants.
Dr. W. Junk, Dordrecht.

Feinsinger, P. and R. K. Colwell. 1978. Community
organization among neotropical nectar-feeding birds.
American Zoologist 18: 779-795.

Fiala, B. and U. Maschwitz. 1 990. Studies on the South
East Asian ant-plant association Crematogaster
bomeensis/Macaranga: adaptations of the ant part-
ner. Insectes Sociaux 37: 2 1 2-23 1 .

Fiala, B. and U. Maschwitz. 1991. Extrafloral nectaries
in the genus Macaranga (Euphorbiaceae) in Malay-
sia: comparative studies of their possible signifi-
cance as predispositions for myrmecophytism. Bio-
logical Journal of the Linnean Society 44: 287-305.

Fiala, B., U. Maschwitz and Y. P. Tho. 1991. The
association between Macaranga trees and ants in
South-east Asia, pp. 263-270. In Huxley, C. R. and
D. F. Cutler, eds.. Ant - Plant Interactions. Oxford
University Press, Oxford.

Fiala, B., U. Maschwitz, Y. P. Tho, and A. J. Helbig.
1989. Studies of a South East Asian ant-plant asso-
ciation: protection of Macaranga trees by
Crematogaster borneensis. Oecologia (Berlin) 79:
463-470.

Folgarait, P. and D. Davidson. 1992. Interspecific and
intraspecific variation in defense investment in Ce-
cropia, p. 1 8. In Abstracts of the Annual Meeting of

62

Journal of Hymenoptera Research

The Association for Tropical Biology , August 9-13,
Honolulu, Hawaii.

Forel, A. 1904. In und mit Pflanzen lebende Ameisen
aus dem Amazonas-Gebiet und aus Peru. Zoologishe
Jahrbiicher, Abteilung fur Systematik, Geographic
und Biologie der Tiere 20: 677-707.

Forel, A. 1928. The Social World of Ants in Relation
with that of Man, 2 vols, trans. Ogden, C. K. G. P.
Putnam’s Sons, London, 551 + 445 pp.

Foreman, D. B. 1978. Myristicaceae (excluding
Horsfieldia), pp. 175-215. In J. S. Womersley (Ed),
Handbook of the Flora of Papua New Guinea, Vol-
ume 1 . Melbourne University Press, Melbourne.

Freese, C. H. 1976. Predation on swollen-thorn acacia
ants by white-faced monkeys, Cebus capucinus.
Biotropica 8: 278-281.

Gautier-Hion, A., J. M. Duplantier, R. Quris, F. Feer, C.
Sourd, J. P. Decoux, G. Dubost, L. Emmons, C.
Erard, H. Hecketsweiler, A. Moungazi, C. Roussilhon,
and J. M. Thiollay. 1985. Fruit characters as a basis
of fruit choice and seed dispersal in a tropical forest
vertebrate community. Oecologia (Berlin) 65: 324-
337.

Gentry, A. H. 1986. Endemism in tropical vs. temperate
plant communities, pp. 153-181. In Soule, M.E.,ed.,
Consemition Biology: The Science of Scarcity ’ and
Diversity. Sinauer Associates, Sunderland, Massa-
chusetts.

Gentry, A. H. 1989. Speciation in tropical forests, pp.
1 1 3-134. In Holm-Nielsen, L. B., I.C. Nielsen and H.
Balslev, eds.. Tropical Forests. Academic Press,
New York.

Gentry, A. H. In press. Diversity and floristic compo-
sition in lowland tropical forest in Africa and South
America. In Goldblatt, P. ed.. Biological Relation-
ships between Africa and South America. Yale
University Press, New Haven.

Gibbs, P. E. and N. Taroda. 1983. Heterostyly in the
Cordia alliodora - C. trichotoma complex in Brazil.
Revista brasileira de Botanica 6: 1-10.

Givnish, T. J. 1987. Comparative studies of leaf form:
assessing the relative roles of selective pressures and
phylogenetic constraints. New Phytologist 106
(Suppl.): 131-160.

Gottsberger, G. 1990. Flowers and beetles in the South
American tropics. Botanica Acta 103:360-365.

Grime, J.P. 1974. Vegetation classification by reference
to strategies. Nature (London) 250: 26-3 1 .

Halle, F., R. A. A. Oldeman, and P. B. Tomlinson. 1978.
Tropical Trees and Forests. Springer-Verlag, Ber-
lin, 441 pp.

Halle, N. 1970. Rubiacees (2e partie), pp. 1-335. In
Aubreville, A. and J.-F. Leroy, eds., Flore du Gabon,

Volume 17. Museum National d'Histoire Naturelle,
Paris.

Harada, A. Y. and W. W. Benson. 1988. Especies de
Azteca (Hymenoptera, Formicidae) especializadas
em Cecropia (Moraceae): distribuigao geografic e
considera^oes ecologicas. Revista brasileira de
Entomologia 32: 423-435.

Hartley, T. G. 1966. A revision of the Malesian species
of Zanthoxylum (Rutaceae). Journal of the Arnold
Arboretum 47:171-221.

Haskins, C. 1939. Of Ants and Men. Prentice-Hall, New
York, 244 pp.

Heithaus, E. R„ T. H. Fleming, and P. A. Opler. 1975.
Foraging patterns and resource utilization in seven
species of bats in a seasonal tropical forest. Ecology
56: 841-854.

Herre, E.A., D. M. Windsor, and R. B. Foster. 1986.
Nesting associations of wasps and ants on lowland
Peruvian ant-plants. Psyche 93: 321-330.

Herrera, C. M 1 984. A study of avian frugivores, bird-
dispersed plants, and their interaction in Mediterra-
nean scrublands. Ecological Monographs 54: 1-23.

Hocking, B. 1970. Insect associations with the swollen
thorn acacias. Transactions of the Royal Entomo-
logical Society of London 122: 211-255.

Holldobler, B. and H. Engel-Siegel. 1984. On the
metapleural glands of ants. Psyche 91: 201-224.

Holldobler, B. and E. O. Wilson. 1990. The Ants.
Belknap Press, Cambridge, Massachusetts, 732 pp.

Howe, H. F. 1984. Constraints on the evolution of
mutualism. American Naturalist 123:764-777.

Howe, H. F. and L. C. Westley. 1988. Ecological
Relationships of Plants and Animals. Oxford Uni-
versity Press, Oxford, 273 pp.

Huxley, C. R. 1986. Evolution of benevolent ant-plant
relationships, pp. 257-282. In Juniper, B. and Sir R.
Southwood, eds. Insects and the Plant Surface. Ed-
ward Arnold, London.

Huxley, C. R. and M. H. P. Jebb. 1991 . The tuberous
epiphytes of the Rubiaceae 1 : a new subtribe - the
Hydnophytinae. Blumea 36: 1-20.

Ibarra-Menrfquez, G. and R. Dirzo. 1990. Plantas
mirmecofilas arboreas de la Estacion de Biologia
"Los Tuxtlas", Veracruz, Mexico. Revista Biologica
Tropical 38:79-82.

Janson, C. 1983. Adaptation of fruit morphology to
dispersal agents in a neotropical forest. Science 2 1 9:
187-189.

Janzen, D. H. 1 966. Coevolution of mutualism between
ants and acacias in central America. Evolution 20:
249-275.

Volume 2, Number 1 , 1 993

63

Janzen, D. H. 1967a. Fire, vegetation structure and the
ant-Acacia interaction in Central America. Ecology
48: 26-35.

Janzen, D. H. 1967b. Why mountain passes are higher
in the tropics. American Naturalist 101: 233-249.

Janzen D.H. 1969. Allelopathy by myrmecophytes: the
ant Azteca as an allelopathic agent of Cecropia.
Ecology 50: 147-153.

Janzen, D. H. 1972. Protection of Barteria
(Passifloraceae) by Pachysima ants
(Pseudomyrmecinae) in a Nigerian rain forest. Ecol-
ogy 53: 885-892.

Janzen, D.H. 1973. Dissolution of mutualism between
Cecropia and its Azteca ants. Biotropica 5: 15-28.

Janzen, D. H. 1973. Sweep samples of tropical foliage
insects: effects of seasons, vegetation types, eleva-
tion, time of day and insularity. Ecology 54: 687-
708.

Janzen, D. H. 1974a. Swollen-thorn acacias of Central
America. Smithsonian Contributions to Botany 13:
1-131.

Janzen, D. H. 1974b. Epiphytic myrmecophytes in
Sarawak: mutualism through the feeding of plants by
ants. Biotropica 6: 237-259.

Janzen, D. H. 1975. Pseitdoinyrmex nigropilosa : a
parasite of a mutualism. Science 188: 936-37.

Janzen, D. H. 1 983. Pseudomynnexferruginea ( Hormiga
del cornizuelo, acacia-ant), pp. 762-764. In Janzen,
D. H. ed., Costa Rican Natural History. University of
Chicago Press, Chicago.

Jebb,M.H.P. 1991. Cavity structure and function in the
tuberous Rubiaceae, pp. 374-389. In Huxley, C. R.
and D. F. Cutler, eds.. Ant - Plant Interactions.
Oxford University Press, Oxford.

Jolivet, P. 1986. Les Fourmis et les Plantes. Paris:
Boubee, 254 pp.

Jolivet, P. 1991 . Ants, plants, and beetles: a triangular
relationship, pp. 397-406. In Huxley, C. R. and D. F.
Cutler, eds ..Ant - Plant Interactions. OxfordUniver-
sity Press, Oxford.

Jordano, P. 1987. Patterns of mutualistic interactions in
pollination and seed dispersal: connectance, depen-
dence asymmetries, and coevolution. American
Naturalist 129: 657-677.

Judd, W. S. 1989. Taxonomic studies in the Miconieae
(Melastomataceae). 111. Cladistic analysis of axil-
lary-flowered taxa. Annals of the Missouri Botanical
Garden 76: 476-495.

Judd, W. S. and J. D. Skean, Jr. 1991. Taxonomic
studies in the Miconieae (Melastomataceae) IV.
Generic realignments among terminal-flowered taxa.
Bulletin of the Florida Museum of Natural History.
Biological Sciences 36: 25-84.

Keigwin, L. D., Jr. 1978. Pliocene closing of the
Isthmus of Panama, based on biostratigraphic evi-
dence from nearby Pacific Ocean and Caribbean Sea
cores. Geology 6: 630-634.

Kiester, A. R., R. Lande, and D. W. Schemske. 1984.
Models of coevolution and speciation in plants and
their pollinators. American Naturalist 124: 220-243.

Koepcke, M. 1972. Ueber die Resistenzformen der
Vogelnester in einem begrenzten Gebiet des
tropischen Regenwaldes in Peru. Journal fiir
Ornithologie 113: 138-160.

Krishna, K. and D. Grimaldi. 1 99 1 . A new fossil species
from the Dominican amber of the living Australian
termite genus Mastotermes (lsoptera:
Mastotermitidae). American Museum Novitates
3021:1-10.

Law, R. and S. Koptur. 1 986. On the evolution of non-
specific mutualism. Biological Journal of the Lin-
nean Society 27: 251-267.

Leenhouts, P. W. and M. Vente. 1982. A taxonomic
revision of Harpullia (Sapindaceae). Blumea 28: 1-
51.

Leston, D. 1 973. The ant mosaic - tropical tree crops and
the limiting of pests and disease. Pest Articles and
News Summaries 1 9: 3 1 1 -336.

Letoumeau, D. K. 1 983. Passive aggression: an alterna-
tive hypothesis for the Piper-Pheidole association.
Oecologia (Berlin) 60: 122-126.

Letourneau, D. K. 1990. Code of ant-plant mutualism
broken by parasite. Science 248: 215-217.

Letourneau, D. K. 1991. Parasitism of ant-plant
mutualisms and the novel case of Piper , pp. 390-396.
In Huxley, C. R. and D.K. Cutler, eds.. Ant - Plant
Interactions. Oxford University Press, Oxford.

Letourneau, D. K., A.G. Feyner, and M. Jebb. 1993.
Coping with enemy-filled space: herbivores on
Endospermum, in Papua New Guinea. Biotropica
25:95-99.

Letouzey, R. and F. White. 1970. Ebenacees, pp. 3-184.
In A. Aubreville and J.-F. Leroy (Eds), Flore du
Cameroon, Volume 11. Paris: Museum National
d’Histoire Naturelle.

Letouzey, R. and F. White. 1976. Chrysobalanacees
nouvelles du Cameroun et du Gabon. Adansonia , ser.
2, 16: 229-243.

Letouzey, R. and F. White. 1978. Chrysobalanacees,
pp. 3-138. In Aubreville, A. and J.-F. Leroy, eds.,
Flore du Cameroun, Volume 20. Museum National
d'Histoire Naturelle, Paris.

Longino, J. T. 1987. Ants in live oak galls in Texas.
National Geographic Research 3: 125-127.

Longino, J. T. 1989a. Geographic variation and com-
munity structure in an ant-plant mutualism: Azteca

Journal of Hymenoptera Research

and Cecropia in Costa Rica. Biotropica 21: 126-
132.

Longino, J. T. 1989b. Taxonomy of the Cecropia -
inhabiting ants in the Azteca alfari species group
(Hymenoptera: Formicidae): evidence fortwo broadly
sympatric species. Natural History Museum of Los
Angeles County, Contribution in Science 412: 1-16.

Longino, J. T. 1991a. Azteca ants in Cecropia trees:
taxonomy, colony structure and behavior, pp. 27 1 -
288. In Huxley, C. R. and D.K. Cutler, eds.. Ant -
Plant Interactions. Oxford University Press, Ox-
ford.

Longino, J. T. 1991b. Taxonomy of the Cecropia-
inhabiting Azteca ants. Journal of Natural History
25: 1571-1602.

Mabberley, D. J. 1979. The species of Chisocheton
(Meliaceae). Bulletin of the British Museum of
Natural History (Botany) 6: 301-386.

Mabberley, D. J. 1985. Florae malesianae praecursores
LXV1I. Meliaceae (divers genera). Blumea 31: 1 29-
152.

Mabberley, D. J. 1987. The Plant-Book. Cambridge
University Press, Cambridge, 707 pp.

Maguire, B. and R. E. Weaver, Jr. 1 975. The Neotropi-
cal genus Tachia (Gentianaceae). Journal of the
Arnold A rboretum 56: 1 03- 1 25 .

Marshall. L. G„ S. D. Webb, J. J. Sepkoski, and D. M.
Raup. 1982. Mammalian evolution and the great
American interchange. Science 215: 1351-1357.

Maschwitz, U. 1974. Vergleichende Untersuchungen
zur Funktion der Ameisenmetathorakaldriise.
Oecologia (Berlin) 16: 303-310.

Maschwitz, U. and B. Fiala. 1992. A new ant-tree from
SE Asia: Zanthoxylum myriacanthum (Rutaceae),
the thorny ivy-rue (Chenkering, Kabu Kabu Utah).
Malayan Nature Journal 46: 1 0 1 - 1 09.

Maschwitz, U„ B. Fiala. Y. F. Lee, V. K. Chey, and F. L.
Tan. 1989. New and little-known myrmecophytic
associations from Bornean rain forests. Malayan
Nature Journal 43: 106-1 15.

Maschwitz, U., B. Fiala, J. MoogandL. G. Saw. 1991.
Two new myrmecophytic associations from the Malay
Peninsula: ants of the genus Cladomyrma
(Formicidae, Camponotinae) as partners of Saraca
thaipingensis (Caesalpiniaceae) and Crypteronia
griffithii (Crypteroniaceae). 1. Colony foundation
and acquisition of trophobionts. Insectes Sociau. y38:
27-35.

McKey, D. 1974. Ant-plants: selective eating of an
unoccupied Barteria by a colobus monkey. Biotropica
6: 269-270.

McKey, D. 1984. Interaction of the ant-plant Leonardoxa

africana (Caesalpiniaceae) with its obligate inhabit-
ants in a rainforest in Cameroon. Biotropica 16: 81-
99.

McKey. D. 1988. Promising new directions in the study
of ant-plant mutualisms, pp. 335-355. In Greuter, W.
and B. Zimmer, eds.. Proceedings of the XIV Interna-
tional Botanical Congress. Koeltz, Konigstein/
Taunus, Germany.

McKey. D. 1989. Interactions between ants and legumi-
nous plants, pp. 673-7 1 8. In Stirton, C. H. and J. L.
Zarucchi, eds.. Advances in Legume Biology, Mono-
graphs in Systematic Botany, No. 29. Missouri
Botanical Garden, St. Louis.

McKey, D. 1991. Phylogenetic analysis of the evolution
of a mutualism: Leonardoxa (Leguminosae:
Caesalpinioideae), pp. 310-334. In Huxley, C. H.
and D. K. Cutler, eds., Ant - Plant Interactions.
Oxford University Press, Oxford.

McKey, D. and D. W. Davidson. In press. Ant - plant
interactions in Africa and the Neotropics: history,
biogeography and diversity. In P. Goldblatt (Ed),
Biological Relationships between Africa and South
America. Yale University Press, New Haven.

McKey, D., P. G. Waterman, C. N. Mbi, J. S. Gartlan and
T. T. Struhsaker. 1978. Phenolic content of vegeta-
tion in two African rain forests: ecological implica-
tions. Science 202: 61-64.

Merlin, P.. J. C. Brackman, D. Daloze, M. Pasteels and
A. Dejean. 1992. New C 26 "/-lactones from the
Dufour’s gland of the urticating ant Tetramorium
aculeatum. Experimenting : 111-113.

Mitter, C. and D.R. Brooks. 1983. Phylogenetic aspects
of coevolution, pp. 65-98. In Futuyma, D. J. and M.
Slatkin, eds.. Coevolution. Sinauer, Sunderland,
Massachusetts.

Moermond,T. C.andJ. S.Denslow. 1985. Neotropical
avian frugivores: patterns of behavior, morphology,
and nutrition with consequences for fruit selection,
pp. 865-897. In Buckley, P. A., M. S. Foster, E. S.
Morton, R. S. Ridgely, and N.G. Smith, eds.. Neotro-
pical Ornithology. American Ornithological Soci-
ety, Washington, D. C.

Monteith, G. B. 1986. Presidential Address. Some
curious insect-plant associations in Queensland.
Queensland Naturalist 26: 105-1 14.

Monteith, G. B. 1987. Endospennum, ants and uraniid
moths in Australia. Queensland Naturalist 28: 35-
41.

Morawetz, W„ M. Henzl and B. Wallnofer. 1992. Tree
killing by herbicide producing ants for the establish-
ment of pure Tococa occidentalis populations in the
Peruvian Amazon. Biodiversity and Conservation 1:

Volume 2, Number 1, 1993

65

19-33.

Nkongmeneck, B.-A. 1985. Le genre Cola au Cameroun.
Annales de la Faculte de Sciences, Biologie-
Biochimie, Universite de Yaounde 3: 5-27.

O'Dowd, D. J. and M. F. Willson. 1989. Leafdomatia
and mites on Australasian plants: ecological and
evolutionary implications. Biological Journal of the
Linnean Society 37: 191-236.

Oliveira, P.S., A. T. Oliveira-Filho and R. Cintra. 1987.
Ant foraging on ant-inhabited Triplaris
(Polygonaceae) in western Brazil: a field experiment
using live termite-baits. Journal of Tropical Ecology
3: 193-200.

Oster, G. F. and E. O. Wilson. 1 978. Caste and Ecology
in the Social Insects. Princeton University Press,
Princeton, New Jersey, 352 pp.

Pennington, T. D. 1981. Meliaceae. Flora Neotropica,
Monograph No. 28. New York Botanical Garden,
New York, 470 pp.

Philipson, W.R. 1984. A synopsis of the Malesian
species of Steganthera (Monimiaceae). Blumea 29:
481-497.

Philipson, W. R. 1985. A synopsis of the Malesian
species of Kibara (Monimiaceae). Blumea 30: 389-
415.

Pontin, A.J. 1958. A preliminary note on the eating of
aphids by ants of the genus Lasius. Entomologist's
Monthly Magazine 94: 9-1 1 .

Pontin, A. J. 1978. The numbers and distribution of
subterranean aphids and their exploitation by the ant
Lasius flavus (Fabr.). Ecological Entomology 3:
203-207.

Prance, G. T. 1972. The Chrysobalanaceae. Flora
Neotropica, Monograph No. 9. New York Botanical
Garden, New York, 410 pp.

Prance, G.T. 1989. Clvysobalanaceae. Flora Neotropica,
Monograph 9S. New York Botanical Garden, New
York, 267 pp.

Putz, F. E. and N. M. Holbrook. 1 988. Further observa-
tions on the dissolution of mutualism between Ce-
cropia and its ants: the Malaysian case. Oikos 53:
121-125.

Riisanen, M.E..J. S. Salo and R. Kalliola. 1987. Fluvial
perturbance in the western Amazon basin: regulation
by long-term sub-Andean tectonics. Science 238:
1398-1401.

Raven, P. H. and D. I. Axelrod. 1 972. Plate tectonics and
Australasian paleobiogeography. Science\l(y. 1379-
1386.

Raven, P. H. and D. 1. Axelrod. 1974. Angiosperm
biogeography and past continental movements. An-
nals of the Missouri Botanical Garden 61: 539-673.

Renner. S. 1986. The neotropical epiphytic
Melastomataceae: phytogeographic patterns, fruit
types and floral biology. Selbyana 9: 104-111.

Ricklefs, R. E. and G. W. Cox. 1972. Taxon cycles in
the West Indian avifauna. American Naturalist 106:
195-219.

Rickson, F. R. 1977. Progressive loss of ant-related
traits of Cecropia peltata on selected Caribbean is-
lands. American Journal of Botany 64: 585-592.

Rickson, F. R. 1979. Absorption of animal tissue
breakdown products into a plant stem — the feeding
of a plant by ants. American Journal of Botany 66:
87-90.

Rickson, F. R. and M. M. Rickson. 1986. Nutrient
acquisition facilitated by litter collection and ant
colonies on two Malaysian palms. Biotropica 18:
337-343.

Rico-Gray, V. andL.B.Thien. 1989. Effect of different
ant species on reproductive fitness of Schomburgkia
tibicinis (Orchidaceae). Oecologia (Berlin) 8 1 : 487-
489.

Ridley, H. N. 1910. Symbiosis of ants and plants.
Annals of Botany 24: 457-483.

Ridsdale, C. E. 1978. A revision of the tribe Naucleeae
s.s. (Rubiaceae). Blumea 24: 307-366.

Risch, S., M. McClure, J. Vandermeer and S. Waltz.
1977. Mutualism between three species of tropical
Piper (Piperaceae) and their ant inhabitants. Ameri-
can Midland Naturalist 98: 433-444.

Risch, S. J. and F. R. Rickson. 1981. Mutualism in
which ants must be present before plants produce
food bodies. Nature (London) 291: 149-50.

Robbrecht, E. 1979. The African genus Tricalvsia A.
Rich. (Rubiaceae-Coffeeae). 1. A revision of the
species of subgenus Empogona. Bulletin du Jardin
Botanique National de Belgique 49:239-360.

Rojo, J. P. 1972. Pterocarpus (Leguminosae-
Papilionaceae) revised for the world.
Phanerogamarum Monographiae 5. Verlag von J.
Cramer, Leutershausen, Germany, 1 19 pp.

Ross, J. H. 1981 . An analysis of the African Acacia
species: their distribution, possible origins, and

relationships. Bothalia 13: 389-413.

Salo, J., R. Kalliola, I. Hakkinen, Y. Makinen, P. Niemela,
M. Puhakka and P. D. Coley, 1 986. River dynamics
and the diversity of Amazon low land forest. Nature
(London) 322: 254-258.

Schemske, D. W. 1983. Limits to specialization and
coevolution in plant-animal mutualisms, pp. 67- 1 09.
In Nitecki, M. H. ed.. Coevolution. University of
Chicago Press, Chicago.

66

Journal of Hymenoptera Research

Schemske, D. W. and C. C. Horvitz 1 988. Plant-animal
interactions and fruit production in a neotropical
herb: a path analysis. Ecology 69: 1 128-1 137.

Schnell, R. 1966. Remarques morphologiques sur les
“myrmecophytes.” Memoires delaSociete Botcmique
de France 1966: 121-132.

Schnell, R. 1967. Contribution a l’etude des genres
Guyano-Amazoniens Tococa Aubl. etMaieta Aubl.
(Melastomacees) et de leurs poches foliaires.
Adansonia 6: 525-532.

Schnell, R., G. Cusset, V. Tchinaye and N. T. Anh. 1 968.
Contribution a l’etude des “acarodomaties.” La
question des aisselles des nervures. Revue Generate
de Botanique 75: 5-64.

Schnell, R. and F. Grout de Beaufort. 1966. Contribu-
tion a l’etude des plantes a myrmecodomaties de
l’Afrique intertropicale. Memoires de VInstitut
Fondamental d'Afrique Noire, No. 75: 1-66.
Melanges Botaniques, Dakar.

Schumann, K. 1 890. Einige weitere Ameisenpflanzen.
Verhandlungen des Botanischen Vereins fur
Brandenburg 31: 113-123.

Schumann, K. and K. Lauterbach. 1901. Die Flora der
deutschen Schutzgebiete in der Siidsee. Berlin.

Schupp, E. and D. H. Feener, Jr. 1991. Phylogeny,
lifeform and habitat dependence of ant-defended
plants in a Panamanian forest, pp. 175-197. In
Huxley, C. R. and D. K. Cutler, eds.. Ant - Plant
Interactions. Oxford University Press, Oxford.

Shattuck, S. O. 1991. Revision of the dolichoderine ant
genus Axinidris (Hymenoptera: Formicidae). Sys-
tematic Entomology 16: 105-120.

Shattuck, S. O. 1992a. A Generic revision of the ant
subfamily Dolichoderinae (Hymenoptera:
Formicidae). Sociobiology 21:1-181.

Shattuck, S. O. 1 992b. Review of the Dolichoderine ant
genus Iridomyrmex Mayr with descriptions of three
new genera (Hymenoptera: Formicidae). Journal of
the Australian Entomological Society 31:13-18.

Simpson, B. B. and C. A. Todzia. 1990. Patterns and
processes in the development of the high Andean
flora. American Journal of Botany 77: 1419-1432.

Snelling, R. R. 1979a. Aphomomyrmex and a related
new genus of arboreal African ants (Hymenoptera:
Formicidae). Contributions in Science from the
Natural History Museum of Los Angeles Countv 3 1 6:
1 - 8 .

Snelling, R. R. 1979b. Three new species of the
palaeotropical arboreal ant genus Cataulacus (Hy-
menoptera: Formicidae). Contributions in Science
from the Natural Histoiy Museum of Los Angeles
County 315: 1-8.

Snow,D.W. 1981. Tropical frugivorous birds and their
food plants: A world survey. Biotropica 13:1-14.

Solereder,H. 1920. Ubereineheterophyllephilippinische
Ameisenpflanze aus der Familie der
Melastomataceae, nebst Bemerkungen iiber das
Auftreten von Amylodextrin-Kornem in den sog.
Perldriisen. Naturwissenschaftliche Wochenschrift
35:689-691.

Spruce, R. 1908. Notes of a Botanist on the Amazon and
Andes. Vol. II. Johnson Reprint Corporation, New
York, 542 pp.

Steenis, C.G. G. J. van. 1967. Miscellaneous Botanical
Notes XVIII. Blumea 15: 145-155.

Steenis, C. G. G. J. van and W. J. J. O. de Wilde, eds.
1950-1989. Flora Malesiana. Kluwer, Dordrecht.

Stevens, P.F. 1978. Meliaceae, pp. 135-174. In
Womersley, J. S. ed.. Handbooks of the Flora of
Papua New Guinea. Volume 1. Melbourne Univer-
sity Press, Melbourne.

Stitz, H. 1910. Westafrikanische Ameisen. 1
Mitteilungen aus dem Zoologischen Museum in Ber-
lin 5: 127-151.

Stitz, 11. 1913. Ameisen und Pflanzen. Die

Naturwissenschaften, 1: 1281-1288.

Stone, B. C., and T. C. Whitmore. 1970. Notes on the
systematy of Solomon Islands’ plants and some of
their New Guinea relatives. XI. Tapeinosperma
(Myrsinaceae). Reinwardtia 8: 3-1 1.

Svoma, E. and W. Morawetz. 1992. Drusenhaare,
Emergenzen und Blattdomatien bei der
Ameisenpflanze Tococa occidentalis (Melasto-
mataceae). Botanische Jahrbuclter 1 14:185-200.

Tennant, L. E. 1989. A new ant-plant, Tetrathylacinm
costaricense, p. 27. Abstracts, Symposium on Inter-
actions between Ants and Plants, Oxford, England,
July 1989.

Terborgh,J. 1983. Five New World Primates. Princeton
University Press, Princeton, New Jersey.

Terron, G. 1970. Recherches morphologiques et
biologiques sur Tetraponera anthracina Santschi et
sur son parasite social temporaire Tetraponera ledouxi
Terron. These, Docteur es Sciences Naturelles,
Universite de Toulouse. 3 1 3 pp.

Thompson, J. N. 1982. Interaction and Coevolution.
Wiley, New York, 179 pp.

Thompson, J. N. 1987. Symbiont-induced speciation.
Biological Journal of the Linnean Society' 32: 385-
393.

Thompson, J. N. 1988. Variation in interspecific inter-
actions. Annual Review of Ecology and Systematics
19:65-87.

Thorne, R. F. 1973. Floristic relationships between

Volume 2, Number 1, 1993

67

tropical Africa and tropical America, pp. 27-47. In
Meggers, B. J., E. S. Ayensu and W. D. Duckworth,
eds.. Tropical Forest Ecosystems in Africa ami South
America. Smithsonian Institution Press, Washing-
ton, D. C.

Tilman, D. 1982. Resource Competition and Commu-
nity Structure. Princeton University Press, Princeton,
New Jersey.

Tschinkel, W. R. 1991. Insect sociometry, a field in
search of data. Insectes Sociaux 38:77-82.

Vasconcelos, H. L. 1990. The relationship between
plant size and ant associates in the ant-plant Maieta
guianensis. Annual Meeting of the Brittish Ecologi-
cal Society (poster session), Manchester, England.

Vasconcelos, H. L. 1991. Mutualism between Maieta
guianensis Aubl., a myrmecophytic melastome, and
one of its ant inhabitants: ant protection against
insect herbivores. Oecologia (Berlin) 87: 295-298.

Vasconcelos, H. In press. Ant colonization of Maieta
guiancensis seedlings, an Amazon ant-plant.
Oecologia.

Vepsalainen, K. and B. Pisarski. 1982. Assembly of
island ant communities. Annales Zoologici Fennici
19: 327-335.

Verdcourt, B. 1 979. A Manual of New Guinea Legumes.
Botany Bulletin, No. 1 1 . Office of Forests, Papua
New Guinea, 645 pp.

Vermeij, G. J. 1978. Biogeography and Adaptation.
Harvard University Press, Cambridge, Massachu-
setts, 332 pp.

Villiers,J.-F. 1984. Le genre Ca/pocfl/v.v (Leguminosae,
Mimosoideae) en Afrique. Adansonia 6: 297-311.

Warburg, O. 1894. Ueber Ameisenpflanzen.
Biologisches Centralblatt 12: 129-142.

Ward, P. S. 1989a. Systematic studies on
pseudomyrmecine ants: revision of the

Pseudomyrmex oculatus and P. subtilissimus species
groups, with taxonomic comments on other species.
Quaestiones Entomologicae 25: 393-468.

Ward, P. S. 1989b. Genetic and social changes associ-
ated with ant speciation, pp. 123-148. In Breed, M.
D. and R. E. Page, eds. The Genetics of Social
Evolution. West view Press, Boulder, Colorado.

Ward, P. S. 1 990. The ant subfamily Pseudomyrmecinae
(Hymenoptera: Formicidae): generic revision and
relationship to other formicids. Systematic Entomol-
ogy 15: 449-489.

Ward, P. S. 1991. Phylogenetic analysis of ant-plant
associations involving pseudomyrmecine ants, pp.
335-352. In Huxley, C. R. and D. K. Cutler eds., Ant
- Plant Interactions. Oxford University Press, Ox-
ford.

Way, M.J. 1954. Studies on the association of the ant
Oecophylla longinoda (Latr.) (Formicidae) with the
scale insect Saissetia zanzibarensis Williams
(Coccidae). Bulletin of Entomological Research 45:
113-134.

Wheeler, W. M. 1920. The subfamilies of Formicidae,
and other taxonomic notes. Psyche 27: 46-55.

Wheeler, W. M. 1929a. The identity of the ant genera
Gesomyrmex Mayr and Dimorphomyrtnex Ernest
Andre. Psyche 36: 1-12.

Wheeler, W. M. 1929b. Note on Gesomyrmex. Psyche
36:91-92.

Wheeler, W.M. 1942. Studies of neotropical ant-plants
and their ants. Bulletin of the Museum of Compara-
tive Zoology 90: 1-262, 57 pis.

Wheeler, W. M. and I. W. Bailey. 1920. The feeding
habits of pseudomyrmine and other ants. Transac-
tions of the American Philosophical Society 22: 235-
279.

Wheeler, W. M. and J. C. Bequaert 1929. Amazonian
myrmecophytes and their ants. Zoologische Anzeiger
82: 10-39.

Whiffin, T. 1972. Observations on some upper Amazo-
nian formicarial Melastomataceae. SIDA 5: 33-41.

White, P. S. 1983. Comer's rules in eastern deciduous
trees: allometry and its implications for the adaptive
architecture of trees. Bulletin of the Torrev Botanical
Club 110: 203-212.

Whitmore, T. C. 1973. Euphorbiaceae, pp. 105-1 13. In
Whitmore, T. C. ed., The Tree Flora of Malaya,
Volume 2. Longman, Kuala Lumpur.

Whitmore, T. C. 1975. Macaranga, pp. 140-159. In
Airy Shaw, H. K. ed., Euphorbiaceae of Borneo.
Kew Bulletin Additional Series 4.

Wilson, E. O. 1 959. Some ecological characteristics of
ants in New Guinea rain forests. Ecology 40: 437-
447.

Wilson, E. O. 1961. The nature of the taxon cycle in the
Melanesian ant fauna. American Naturalist 95: 169-
193.

Wilson, E. O. 1985. Invasion and extinction in the West
Indian ant fauna: evidence from the Dominican am-
ber. Science 229: 265-267.

Wilson, E. O. 1987. The arboreal ant fauna of Peruvian
Amazon forests: a first assessment. Biotropica 19:
245-251.

Young, B. E., M. Kaspari and T. E. Martin. 1990.
Species-specific nest site selection by birds in ant-
acacia trees. Biotropica 22: 310-315.

Zizka, G., U. Maschwitz and B. Fiala. 1990. Pflanzen
und Ameisen: Partnerschaft furs Uberleben.
Palmengarten 15: 1-126.

68

Journal of Hymenoptera Research

Appendix 1 . Summary of ants living regularly in symbiotic association with one or more
host species. Questionable or massing data are indicated by question marks. Columns:

(1) Ant genera (superscript indicates carton-building typical of the genus, though not
necessarily of plant-ant species) in biogeographic regions (N) = Neotropical, (E) = Ethiopian,
(M) = Malagasy, (O) = Oriental, and (A) = Australian regions.

(2) Host taxa have growth forms: T = tree; U = treelet or understory tree; S = shrub; L =
liana or vine, R = rattan, B = bamboo and H = hemiepiphyte. Habitats include: b = mountain
brooks; e = edge, second growth, riparian environments; g = forest light gaps; 1 = littoral
scrub; p = primary forests; s = savannahs or dry forest, and a = aguajals or swamps.

(3) Ants nest in domatia comprised of: L = leaf pouches; S = naturally hollow stems; Sp =
pithy stems, hollowed by ants; I = swollen intemodes; P = swollen petioles or bases of
petioles; Ps = petiolar sheath; R = swollen rachi and petioles; St = persistent stipules (in-
flated or folded); Sh = persistent spathe; Th = swollen thorns; F = swollen flowering shoots;
G = gall-like swellings; C = carton shelters around domatia, folded leaves, and/or hairs or
spines; Cl = cavity formed by leaf base clasping stem; B = insect borings; O = inflated ocrea
(proximal extension of leaf sheath beyond the petiole), A = erect, narrow auricles on each
side of petiole, at the terminus of the sheath; Ac = acanthophylls, or basal pinnae reflexed
backward to form a secluded cavity at the base of a palm frond; Ga = galleries enclosed by
interlocking combs of spines, forming collars on leaf sheaths, and T = vast chambers exca-
vated inside tree trunks by ants and partitioned by carton. Plant pubescence: y = domatia and
stems bear long, dense hairs or spines, likely to inhibit movements of larger bodied ants; n =
such hairs or spines lacking, or s = only a subset of plants have these hairs.

(4) Ants prune vines and vegetation around their hosts: Y = obligate for plant-ants in this
genus; S = in at least some ant associates of the host genus; F = where known, pruning is
facultative, i.e., in the presence of enemy ants; N = not yet reported for the ant genus on this
host genus. Host fidelity (foraging predominantly or entirely on the host): y = yes; n = no; i
= for young (incipient) but not established colonies.

(5) Food types include: P = pearl bodies; B = other specialized food bodies; H = exudates
and bodies of homoptera (Coccoidea); E = extrafloral nectar; N = floral nectar; G =
uncharacterized exudates of tiny glands; F = fungi; VV = lipid-rich and/or protein-rich plant
wounds, or heteroplasias caused by traumatic injury by ants; O = pollen; T = glandular
trichome.

(6) Plants have evolved apparently specialized structures to house ants: Y = yes; N = no.

(7) Estimated number of congeneric ant species found regularly on the host genus; prob-
ability of more (+) or several more (++) indicated parenthetically. Square brackets denote
ants known to be unspecialized, or whose specialization is in doubt.

(8) References for data on ants or plants: Bq = Bequaert 1922; W = Wheeler 1942; S&B
= Schnell and Beaufort 1966; B = Benson 1985; H = Huxley 1986; J = Jolivet 1986; H&W
= Holldobler and Wilson 1990; I) = Davidson et al. 1989; IP = in press; PC = personal
communication, I)D and DM = respective author’s observations.

Volume 2 , Number 1 , 1993

^ Z Z

o\

& o c>

z § §
§ I

S £

a*|

« Q ^

c

<

a.

o-

oj

%

co

ro

ro

CJ

g

s

<u

o.

CO

2

>-

><

o»

z

>

Y?

5

~o

8

a>

Oh

>-»

Cu

IE

O'

PJ

ffi

O'

u-

H

DQ

o*

oq

CQ

pq

c

*2

pG

>»

>»

O'

io

>>

w

<

H

CO

o'

o'

o'

o'

co"

or

c

q

m

a>

q

q

c

O.

d,

xf

q

' '

Z

_c

co

GO

o'

co

H

CO

rt _

x ^2

c3 S

H *

X z

< H

pq

<

UJ

U

< “

§ I

w &
u u

pq «

< %
z ^3
a c

2 1,2

Sf ~

CLh CL

w

<

w

u

s £

I?

u 2

pq

s

oS

>-

o

Q
D

ffl r
CO Z 1
cu ^

pq

<

w

u

<

a V Or S

UJ

<

UJ

u

<

CQ

§• <

G a to ^

■g 2 ^ Z

z z z z

Platymiscium T? ?,? ?,? H? N 2(+) " + W

Pterocarpus rea FR,n ?,? ? Y [1] " + Forel 1904; Spruce 1908; Rojo 1972

*Tachigali T ' m P,n S,y H Y 9(+)

Pleurothyrium Vp Sp,n ?,? H N 1

POLYGONACEAE

Ruprechtia u? S,n ?,? H? Y? [1] "+ Spruce 1908; B; Brandbyge 1990

Journal of Hymenoptera Research

Q &

. . T3

g I

+ *

= + Q

g £

bJO ^

I U

O 3

K <

£ -§

!> E— i Q D-

£

(X

z

Z

Z

Z

Z

pq

£

c^-

O

a

pq

pq

pq

pp

X

X

X

X

X

X

X

X

»•

0‘

o-

o-

>,

O’

>•.

>»

O’

z

o'

>

a

c

c

c

c

c

c

d

d

5

x:

d

d

d

CO

CO

H

CO

CC

CO

CO

&

W

< P P

W 5 s
y S 5 pq
^ c c <

7 ^ u

o & £ y

o - «■ 03

a ^ <

ffl ^ ^ u.

pq

<

tq

« U

*„ <

§ d
^ O 8-

LU
<
PP
U

^ ^ ^ w <

"c 3 o E 7? P

§ ? c 53 :

£ c/3 55 <—i

£ -2P < | <

o u at <§ 55

Z ‘5

w
<
- w

3 3

S' £
■S pq
S ffl
§ 0*
s' W

o >

Z Z Z Z S' W o

w O

(E) Vitex Le Sp,n ?,y FW N 1 " + Bq; J

MYRM1CINAE
Tribe Cephalotini

c Zacryptocerus BORAG1NACEAE

(N) Cordial I,n N,y ? Y 1 Carroll 1983; W; J. Longino PC

Volume 2, Number 1, 1993

>■

d

Q

OQ

o\ -h

ON , 03

rl ri*

t~" °o

73 2 P

<U g

so g X

S3 g X

PQ Qu

da da ^

DQ 22

* g

■8 I

O 3
42 XJ

> 00

da da

X

da

PQ

^ da

rf — —

>H >H

X

DU

X

x x o

x

<

x

u

<

DQ

<

x

<

x

x- u
73 <C
c Z
O K

■s o
’> <
2 £
a o

.72 DQ

<

PJ

u

<

Du

O

OP

u

PJ

u

>, UJ
DU <
PJ
U
<
z
<
x
<
oa
O

x

<

x

u

<

H

<

S

o

H

00 q &Q <3 ~

QP
X

u

.£ x
=5 S

X

< .

X <??

U

< .2

5 §

§ d

x

<

X

u

<

2

x

z

X

<

X

u

<

a £ CQ £

55 X 2 D = ^

ad O oa Dip Q ^

| Z

Z Z

Z Z ~ Z z z § o

H ^

Crematogastrini

Crematogastef 1 CECROPIACEAE

(N) Cecropia Jg S,n N,y BPH Y 1 DD

(E) Musanga * e S,n N,n H N [9+] Duviard & Segeren 1974

EUPHORB I ACE AE

Macaranga Vpe St,n N,n E(P?)

Journal of Hymenoptera Research

Q

Q

DQ

Q

13

3 2 S S S

X Q Q Q Q

— ON
ON ^

" s n

M m

Q S >

— NO

— fN ' — i

>- >- Z >- ^ >-

gj pj DC DC

ac dc o- m a w

a a

a a
Z Z

d H

- M MH C/3

w

w
u
<

5

C>

-Cj

c-> <-> c- c: ^ a

^ t ^ ^ ^

55 UJ
<
ffl

u
<
H
<

PU
<
W

J J

g

on

I 3

w
<
w
u
; <

; d
1 s

w O.

w o o o w o

z. z

Aphanamixis Up Sp,n ?,? H N [1+] Warburg 1894; Mabberley 1985;

Volume 2 , Number 1 , 1993

0Q CQ
co co

CD 03
co w

s s

Q Q

CD —

« S I

Q OC

CQ

ca

CQ CQ

=a da

on co

, CQ
cr cr da cr a

00 CQ CO CQ CQ

s

a

CQ

ca

CQ

^ CQ

>- >- > >- z >< >< >< z >< >*

a UJ

£ %

£ £ a

EX X X X o- X X X X X

^ z,

LU

<

PP

u
<

5
x s
ca u

G o ^

L L ~ -L

c: *■ 2: ~c

O'- -r= ? tl

= S

PQ
<
U3
U

§ ES

I 5

^ oc

PJ

< 3

PQ ss

UP

<

pp ^

U -Q <

, z

H -§ W ^

a. ^ c* S
< ^ pp -2
co Q > U

“2 £

m

<

pp

u

<

£ C-.

o 3
< &
O 5

CQ U

a 2 w pp w o S'owS'w

o -§ e
T3 -S Z

GESNERIACEAE
Besleria Sp L,

MELASTOMATACEAE

e Clidemia s.s. Se L,y N,y G(H?) Y 1(+) B; D

Conostegia Sp L,y ?,y ? Y 1 ,[ 1] B; Judd & Skean 1991; L. Tennant

Journal of Hymenoptera Research

un

On vO

8

£ Ss

> %

Q £q

I . - <u

CQ I PQ

« e

cq j

2 gj p£

S_ ' ' ON

w> £ ~

3 ^ <=

CQ £ =3

Q

£

CQ

-H ON

c —

s s

Q Q

— , <N

o — —

Cu

O

CQ

E

ffi DC

2 PL)

i w

.£ PC <4
X W ^

^ 3 ■&

^ Dh 0,

W =

< 5.

CL) CS

< a

3 §

< <5J

Pi kJ

< J

W

u

<

H

<

S

o u

H p-

CQ S

S b

w

w

< s <

— pq

1 ^

.S- CQ

a s

w

<

w

u

<

3

D

DC

pq

<

w

u

<

2 2

§

£ 2

< Cl,
<1

Q co

2 5 3

is 2

5 «

u w o

a.

Volume 2 , Number 1 , 1993

Q

Q

O

Q

Q

Q

d

2

Q

S 1

Q u

ffl I

Q

Q

=3 2

g a

Q CQ

+ + — ,

^ >-

Z Z

Z >~

£

X X

>• p

c c

z z

p

<

p

u

<

3

D

P

p

<

p

u

<

oa

<

p

p

u

<

3

p

P

P

<

P

u

£

<

S

o

H
co
<
P
. P

CO CQ P P

o o

■2 ~

< § u wo szg.fi q<<

I

p
<
p
* U

15

"S H
~ co

§D 3

& >*

p

<

p

u

g

. p

<r *

< 3

S §' S

P p
1 <
Sc j

£ <

Co P

S. B. P
P * U
© -S <

Q k

o o

5 P

fi« <

a P

IS

Cj rO
•2 <2
§ 2

< 23 p

ON A
ON

Journal of Hymenoptera Research

oo

On

13 On

fc 2: Q

3 8«

£ Q £

X

01) o

C ON

Q

Q

X ®
X ^

t!

<u

£

X

Q

* Q
£ «
DQ ^

Q

£

03

=3

Q

Q

<N —1

+ + + X,

£ >H

>H >H

^ ^ ^

a

x

x

c^-

C-. a e-

c^* c »

2 2

X

<

X

u

<

2

a

<

x

o

X

X

<

X

u

<

X

o

X

u

X

u

X

<

X

u

<

i |

£ 0

S' 00

§ >-
s *

© X

u u

X

<

X

u

<

Sf H

u tS Oi

*5 5

5 6« U

2 5 U

SK 6j <
- L? X

O b x

W X
<
X

u
<
D. £— 1
'=. <
I s
I.R

■2 <3 :

~ 6u

5 2

°°

x

<

X

u

<

2

o

a

>

X

o

X

X

* 2
s* U

g < .a -3

X

’ D s
X Q

,2 : 2

5 -S'

Q 0-

2 2

2 2

2 2

Volume 2 , Number 1 , 1993

U

Cl

U 2

C S
« h

Di

u

£

J

u

s =3

Q

Q

QQ

s

s

Q

Q

Q

Q

cr

cr

cr

JJ

CQ

CQ

QQ

Q

U

> >-

PJ

tu x

r co

pg 3 d
<2-2
PJ .3

y ^ 1

ffl 5 :

u- ^j as

— ,

—

J

2

+

2

1 1

1 1

of

o*

o-

>

JH

>-

>-

z

><

>

>-

>-

>-

PJ

X

PJ

PJ

PJ

X

0 *

X

O'

O*

O’

c

O’

e

o-

O’

o-

o-

O’

or

Z

z

o'

Z

o'

0 '

0 '

or

o^

c

>>

y

q

q

q

>>

q

yj

c

>>

J

oo

PJ

J

J

<£

00

PJ

pj

PJ

3

O.

c.

U

PJ

<

<

PJ

u

Cl

D

o

S 1

PJ “|

3

3'

|

<!

y

H

a,

3’

S

<

PJ

u

-3,

CL

3

pj

u

<

g

<

z

PJ

c

1

< ^
CQ

"5

"2

2

3

CO

5

.Cj

■H

l!

<

3

| — »

1

'kj

•2

g

■2

3

2

0

CL

1

L. ~ ^

tL ^ 5: -J

S S' S <3 0 C

^ SiJ

co Q

PJ

<

PJ

u

<

J

X JS.

U g

g 3=

ts a

w O s O w

S*

K K

< o w < u a a u pj

(E) Scaphopetalum SUp L,y ?,? ? Y 1 Bq; Dejean et al. 1990; DM

FORMICINAE
Myrmelachistini
Aphomomyrmex FABACEAE

78

Journal of Hymenoptera Research

s

Q

— o

ON —

00 ON
ON ON

N N

‘I £

-C -C

O CJ
C3 c§

s s

D X

Q

Q

Q

Q

.2 so

£

cd

U

— , <u

o ^

=s DC

cr ..
■C d

O Orj

£ £ o.

'5b ’5b

N SI
<u aj

g £
2 «
o o

s s

■a i

O Oh ON

u

Oh

£ -

i i

ro ^

>H

z z

x

X

x

X X

X

o

X

o

r^r >h

X

<

X

u

<

z

x

CQ

2

x

>

x

<

x

u

<

z

o

OS

X

>*

OS

u

x X C/2 c/o

K- X
■§ < ..

§ U "s

IS I

c- < £

X X ^

<

X

, u

la

3 §

X

<

X

u

<

z

o

<

CS

o

PQ

X

<

X

u

<

E- 1

<

s

O -
H

- rS
X O

< § I

X <j o

X x o

m ? ^

o o.

S I S

z z

z z

MELIACEAE

(N) Guarea Tp Sp,n N,y H N 2(+)[l] J. Longino PC; Pennington 1981

RUBIACEAE

(N) Duroia Upe l,y Y,i' (G?)H Y 1(+) Campbell et al. 1989; DD

Petalomyrmex FABACEAE

Volume 2 , Number 1 , 1993

oo —

Q

Q

S

H

s

Q

£

a

Q

dd

n —i

a

33

X

Oh

CQ

'z, C

a

<

a

u

<

a

D «

u ^

uu -2
H

oo <J

£ <

a, a

u a

§ <

Jg.

A «S < S

5 3 M ^

^ < Z

5, -o O

P g 2

Ho

c m

<u

a

*5- <
a a

<
a, a

eP D

r a

a

< 5
3 a s
5 U g
~ < -5

£

a
<

u §' a .g S o

< is cq s a:

5 ■§ £ s ^ u

• D 5 a -Si s a

t* r^j > a ^ u

a
<
a
u
<
5 S

s o
? x

G £:
su a
u a

"o.

<u

-2 a

'§>,f w

a a

u a

o s.

Endospermum Tpe Sp.n S,y EW Y 1 Bq; J; DD, Monteith 1987;

Letoumeau et. al. PC

Macaranga Tes S,n ?,? ? Y 2 Fiala & al. 1990; F. Rickson PC

PALMAE

f Daemonorops Rjp Ga,? ?,? H N? 4 Ridley 1910; DD; Rickson &

Rickson 1986; Dransfield &
Manokaran 1978.

f Korthalsia Rp 0,s N,? H Y 3(+) B; H&W; DD; Dransfield 1984

RUBIACEAE

Journal of Hymenoptera Research

s £

£ £
X XJ

o o

c/3 c/3

e3 cs

s s

=3 o y
£ £

Q Q

£

£

>< >«

z z

z z

PL) ed

Z « H "a

5 ° ~ ^ ~

<

6

a

o z

O CO

O O

pq
<
w
u
<
5

s I

< y

PJ
<
W

u
<
z
<

<
_ CO

O
~ co

5

Si oc

5 ®
co U

< O <

Q ^

pq

<

pq

U

£ J

■IS

•2 <

« y

Q

Dichapetalum L1 L,y ?,? E? Y ? Breteler 1986 & PC

ELAEOCARPACEAE

Elaeocarpus TUp S,? ?,? ? ? ? J;Coodel981

EUPHORBIACEAE

Volume 2, Number 1, 1993

U

0-

£ J3 fi JJ £

Q

Q

£

X

o

<

” - U

£

ac

C-. >< >■< >-!

-e: Q S -2

O ,

Q O

G ~ ^ Si

w r- 2“

S ^ s;

w

2

!z

rjn ^ *

O g,<5
< •= f- -S

5 1 § |

£ ^ o £

w

<

:• w

■q U
C <

5 2
a <

c o

u J

53 J

W

u
<
t— 1
<

S Sr

O 2
H
uo

, >> »•
J CU

^ <
C J

a w

£ S

w

<

w

u

<

5 £ S

< < < 2 <

z Z Z o z

MONIMIACEAE

Kibara Vp I,n ?,? H Y? ? Bq; J; Philipson 1985

NYCTAGINACEAE

Pisonia 1 S,y ?,? ? ? ? H&W

PALMAE

Journal of Hymenoptera Research

a

I

os OS

tu <u <u a> <u ->3

x x 75 x x x 71

C /5 t /5 D V 5 V 5 1/3 U

c c x c e c is

2 2 c 2 2 2 e

, Q Q 2 Q 5 Q 2

U

CL

x 71

c<i cr
m X

O S'*
2 X

& ffl
O’ ’3

ffl I Cfl

1 ?
>» §
I f

i 5

z >- z

c- X X

uooouo<<

in x in c /3 00

%

~5 E ?

a J 3 a J 3 ~ 2

r c a a a E w

U U U X X r s 3

- X

~ <

5 X
■S U
S- o <

1 |w . ..

& o S: -S- D

x S' x X o'

x

<

X

u
<
£ x

^ a ?*■> .si

5 s

w IU

•|S

fc u

s <£

00 E_

1 X
£ X

X

H <

=L X

5 u
-5 <
£‘ Q
2 Z

^ CL

j 5 <

N m

X X

<c

X

u
<
2
<
p. X
3. X
■S X

X X

IS

m

X

<

X

u
<
X
X
j o
•S X
^ <
■S? §
^ in

X

X <

< W

X x

u <

< * W

U 3

3 1 d

S 5 s

I IS

on co‘ H

OOCOXXXO

X X < < X X

Volume 2, Number 1, 1993

E

! ~o

£ E

<U o is

(U c
X) g
Xi O
rt ^3

§ §

rr> q

OS* o £ >» g .

* a ■

o

w

UP

<

tu

u

<

w

CQ

C*

pj

>

O PS

c S

rt C

<u p3

X <u

W» §

o flj 3

^ 2 — os

O c3 C O O

— PS
:0 C/3

ffi X)

S S jj u]

a t 3 iu .5

E >. jj

g S G

I §
1 1 1 1 1 ■
i I S ■§ -g •

aW o m £ ,

S £

2 -a

C ^ >>

s s i

S ps

E c/3

2 cu £

« Mg ^

.E '-> ps
s o
=3 O' O
<-> 00 K
‘S

' — cj ?3 jj <L> S3

! « ^ |

L*

J fi-S E ’

g>

o 3

X E x

C o <u

_ « -a ==
2 | " I

•N ^ z

5 I

E

£ ON

* %
<u

op TO

: 3 ■
! X

1 e SI 3, — 2

r Cd _ ~

CJ , 1

3 £

: © 2

' e3 =°

x 3 C

O T3

J o< S '

£ ^ O

■s i 1 “

O'

a .2 *

v* 2

5 5

— ■ <u 3

e2 I

R ° '

.E I!

2 £ £ '

i 1 I § I

W Q "ot • •£

I ~ S* 9- -a

2

2 ii

qj w

<r 5

oj C

E £

O n
on ^

^ x

"1 2

§ X £ 3 ;

a is s»

o

2 x

■S 1|

1 1 1

O d <2

:l 5 I

S q o .

■a ■§ X :
X 2 *

"a, '£ .2 .

o ~

3 >, ;= £ 2 s

( s r - - j

E <

1) —

~ < :

£

£

<